From 04abf66d78420c410da9bf9e4ba82d88223f95b1 Mon Sep 17 00:00:00 2001
From: Chris Basoglu <cbasoglu@microsoft.com>
Date: Fri, 25 Mar 2016 17:31:02 -0700
Subject: [PATCH] Prepend timestamp to log lines when tracing flag is turned on

---
 .../03_ResNet_Train_AddBNEval_Test.log.160    | 12828 +++++-----
 .../04_ResNet_56_Train_AddBNEval_Test.log.160 | 19804 ++++++++--------
 Source/CNTK/CNTK.cpp                          |   157 +-
 Source/Common/Include/ProgressTracing.h       |    36 +-
 Source/Common/fileutil.cpp                    |     5 -
 .../ComputationNetworkAnalysis.cpp            |     4 +-
 .../ComputationNetworkEvaluation.cpp          |    23 +-
 Source/SGDLib/SGD.cpp                         |   191 +-
 Source/SGDLib/SGD.h                           |     2 +-
 .../DoublePrecision/baseline.cpu.txt          |     8 +-
 .../DoublePrecision/baseline.gpu.txt          |     8 +-
 .../DoublePrecision/baseline.windows.cpu.txt  |     8 +-
 .../DoublePrecision/baseline.windows.gpu.txt  |     8 +-
 .../SinglePrecision/baseline.cpu.txt          |     8 +-
 .../SinglePrecision/baseline.gpu.txt          |     8 +-
 .../SinglePrecision/baseline.windows.cpu.txt  |     8 +-
 .../SinglePrecision/baseline.windows.gpu.txt  |     8 +-
 .../01_OneHidden/baseline.linux.debug.cpu.txt |     2 +-
 .../01_OneHidden/baseline.linux.debug.gpu.txt |     2 +-
 .../baseline.linux.release.cpu.txt            |     2 +-
 .../baseline.linux.release.gpu.txt            |     2 +-
 .../baseline.windows.debug.cpu.txt            |     2 +-
 .../baseline.windows.debug.gpu.txt            |     2 +-
 .../baseline.windows.release.cpu.txt          |     2 +-
 .../baseline.windows.release.gpu.txt          |     2 +-
 .../Image/MNIST/01_OneHidden/testcases.yml    |     2 +-
 .../baseline.linux.debug.cpu.txt              |     2 +-
 .../baseline.linux.debug.gpu.txt              |     2 +-
 .../baseline.linux.release.cpu.txt            |     2 +-
 .../baseline.linux.release.gpu.txt            |     2 +-
 .../baseline.windows.debug.cpu.txt            |     2 +-
 .../baseline.windows.debug.gpu.txt            |     2 +-
 .../baseline.windows.release.cpu.txt          |     2 +-
 .../baseline.windows.release.gpu.txt          |     2 +-
 .../Image/MNIST/02_Convolution/testcases.yml  |     2 +-
 .../baseline.linux.debug.cpu.txt              |     2 +-
 .../baseline.linux.debug.gpu.txt              |     2 +-
 .../baseline.linux.release.cpu.txt            |     2 +-
 .../baseline.linux.release.gpu.txt            |     2 +-
 .../baseline.windows.debug.cpu.txt            |     2 +-
 .../baseline.windows.debug.gpu.txt            |     2 +-
 .../baseline.windows.release.cpu.txt          |     2 +-
 .../baseline.windows.release.gpu.txt          |     2 +-
 .../MNIST/03_ConvBatchNorm/testcases.yml      |     2 +-
 .../MultiGpu/baseline.linux.debug.cpu.txt     |     2 +-
 .../MultiGpu/baseline.linux.debug.gpu.txt     |     2 +-
 .../MultiGpu/baseline.linux.release.cpu.txt   |     2 +-
 .../MultiGpu/baseline.linux.release.gpu.txt   |     2 +-
 .../MultiGpu/baseline.windows.debug.cpu.txt   |     2 +-
 .../MultiGpu/baseline.windows.debug.gpu.txt   |     2 +-
 .../MultiGpu/baseline.windows.release.cpu.txt |     2 +-
 .../MultiGpu/baseline.windows.release.gpu.txt |     2 +-
 .../Other/Simple2d/MultiGpu/testcases.yml     |     2 +-
 .../Simple/baseline.linux.debug.cpu.txt       |     2 +-
 .../Simple/baseline.linux.debug.gpu.txt       |     2 +-
 .../Simple/baseline.linux.release.cpu.txt     |     2 +-
 .../Simple/baseline.linux.release.gpu.txt     |     2 +-
 .../Simple/baseline.windows.debug.cpu.txt     |     2 +-
 .../Simple/baseline.windows.debug.gpu.txt     |     2 +-
 .../Simple/baseline.windows.release.cpu.txt   |     2 +-
 .../Simple/baseline.windows.release.gpu.txt   |     2 +-
 .../Other/Simple2d/Simple/testcases.yml       |     2 +-
 .../FeedForward/baseline.linux.debug.cpu.txt  |     2 +-
 .../FeedForward/baseline.linux.debug.gpu.txt  |     2 +-
 .../baseline.linux.release.cpu.txt            |     2 +-
 .../baseline.linux.release.gpu.txt            |     2 +-
 .../baseline.windows.debug.cpu.txt            |     2 +-
 .../baseline.windows.debug.gpu.txt            |     2 +-
 .../baseline.windows.release.cpu.txt          |     2 +-
 .../baseline.windows.release.gpu.txt          |     2 +-
 .../Speech/AN4/FeedForward/testcases.yml      |     2 +-
 .../AN4/LSTM/baseline.linux.debug.cpu.txt     |     2 +-
 .../AN4/LSTM/baseline.linux.debug.gpu.txt     |     2 +-
 .../AN4/LSTM/baseline.linux.release.cpu.txt   |     2 +-
 .../AN4/LSTM/baseline.linux.release.gpu.txt   |     2 +-
 .../AN4/LSTM/baseline.windows.debug.cpu.txt   |     2 +-
 .../AN4/LSTM/baseline.windows.debug.gpu.txt   |     2 +-
 .../AN4/LSTM/baseline.windows.release.cpu.txt |     2 +-
 .../AN4/LSTM/baseline.windows.release.gpu.txt |     2 +-
 .../Examples/Speech/AN4/LSTM/testcases.yml    |     2 +-
 .../RNN/baseline.linux.debug.cpu.txt          |     2 +-
 .../RNN/baseline.linux.debug.gpu.txt          |     2 +-
 .../RNN/baseline.linux.release.cpu.txt        |     2 +-
 .../RNN/baseline.linux.release.gpu.txt        |     2 +-
 .../RNN/baseline.windows.debug.cpu.txt        |     2 +-
 .../RNN/baseline.windows.debug.gpu.txt        |     2 +-
 .../RNN/baseline.windows.release.cpu.txt      |     2 +-
 .../RNN/baseline.windows.release.gpu.txt      |     2 +-
 .../Text/PennTreebank/RNN/testcases.yml       |     2 +-
 .../AlexNet/baseline.linux.debug.gpu.txt      |     2 +-
 .../AlexNet/baseline.linux.release.gpu.txt    |     4 +-
 .../AlexNet/baseline.windows.debug.gpu.txt    |     2 +-
 .../AlexNet/baseline.windows.release.gpu.txt  |     2 +-
 .../EndToEndTests/Image/AlexNet/testcases.yml |     8 +-
 .../QuickE2E/baseline.linux.debug.gpu.txt     |     4 +-
 .../QuickE2E/baseline.linux.release.gpu.txt   |     4 +-
 .../QuickE2E/baseline.windows.debug.cpu.txt   |     4 +-
 .../QuickE2E/baseline.windows.debug.gpu.txt   |     4 +-
 .../QuickE2E/baseline.windows.release.cpu.txt |     4 +-
 .../QuickE2E/baseline.windows.release.gpu.txt |     4 +-
 .../Image/QuickE2E/testcases.yml              |     8 +-
 .../ModelExport/Model0/baseline.txt           |     2 +-
 .../ModelExport/Model0/testcases.yml          |     2 +-
 .../ModelExport/Model1/baseline.txt           |     2 +-
 .../ModelExport/Model1/testcases.yml          |     2 +-
 .../DoublePrecision/baseline.cpu.txt          |     8 +-
 .../DoublePrecision/baseline.gpu.txt          |     8 +-
 .../DoublePrecision/baseline.windows.cpu.txt  |     8 +-
 .../DoublePrecision/baseline.windows.gpu.txt  |     8 +-
 .../SinglePrecision/baseline.cpu.txt          |     8 +-
 .../SinglePrecision/baseline.gpu.txt          |     8 +-
 .../SinglePrecision/baseline.windows.cpu.txt  |     8 +-
 .../SinglePrecision/baseline.windows.gpu.txt  |     8 +-
 .../baseline.cpu.txt                          |     2 +-
 .../baseline.gpu.txt                          |     2 +-
 .../baseline.windows.cpu.txt                  |     2 +-
 .../baseline.windows.gpu.txt                  |     2 +-
 .../DiscriminativePreTraining/testcases.yml   |     8 +-
 .../Parallel1BitQuantization/baseline.cpu.txt |     6 +-
 .../Parallel1BitQuantization/baseline.gpu.txt |     6 +-
 .../baseline.windows.cpu.txt                  |     6 +-
 .../baseline.windows.gpu.txt                  |     6 +-
 .../baseline.cpu.txt                          |     6 +-
 .../baseline.gpu.txt                          |     6 +-
 .../baseline.windows.cpu.txt                  |     6 +-
 .../baseline.windows.gpu.txt                  |     6 +-
 .../ParallelCrossValidation/baseline.cpu.txt  |     4 +-
 .../ParallelCrossValidation/baseline.gpu.txt  |     4 +-
 .../baseline.windows.cpu.txt                  |     4 +-
 .../baseline.windows.gpu.txt                  |     4 +-
 .../ParallelNoQuantization/baseline.cpu.txt   |     6 +-
 .../ParallelNoQuantization/baseline.gpu.txt   |     6 +-
 .../baseline.windows.cpu.txt                  |     6 +-
 .../baseline.windows.gpu.txt                  |     6 +-
 .../baseline.cpu.txt                          |     6 +-
 .../baseline.gpu.txt                          |     6 +-
 .../baseline.windows.cpu.txt                  |     6 +-
 .../baseline.windows.gpu.txt                  |     6 +-
 .../DNN/SequenceTraining/baseline.gpu.txt     |     2 +-
 .../SequenceTraining/baseline.windows.gpu.txt |     2 +-
 .../Speech/DNN/SequenceTraining/testcases.yml |     8 +-
 .../Speech/DNN/WriteCommand/baseline.cpu.txt  |     2 +-
 .../Speech/DNN/WriteCommand/baseline.gpu.txt  |     2 +-
 .../DNN/WriteCommand/baseline.windows.cpu.txt |     2 +-
 .../DNN/WriteCommand/baseline.windows.gpu.txt |     2 +-
 .../Speech/DNN/WriteCommand/testcases.yml     |     8 +-
 .../baseline.cpu.txt                          |     2 +-
 .../baseline.gpu.txt                          |     2 +-
 .../baseline.windows.cpu.txt                  |     2 +-
 .../baseline.windows.gpu.txt                  |     2 +-
 .../DNN/DiscriminativePreTraining/run-test    |     6 -
 .../DiscriminativePreTraining/testcases.yml   |     8 +-
 .../Parallel1BitQuantization/baseline.cpu.txt |     6 +-
 .../Parallel1BitQuantization/baseline.gpu.txt |     6 +-
 .../baseline.windows.cpu.txt                  |     6 +-
 .../baseline.windows.gpu.txt                  |     6 +-
 .../DNN/Parallel1BitQuantization/run-test     |     6 -
 .../baseline.cpu.txt                          |     6 +-
 .../baseline.gpu.txt                          |     6 +-
 .../baseline.windows.cpu.txt                  |     6 +-
 .../baseline.windows.gpu.txt                  |     6 +-
 .../run-test                                  |     6 -
 .../ParallelNoQuantization/baseline.cpu.txt   |     6 +-
 .../ParallelNoQuantization/baseline.gpu.txt   |     6 +-
 .../baseline.windows.cpu.txt                  |     6 +-
 .../baseline.windows.gpu.txt                  |     6 +-
 .../DNN/ParallelNoQuantization/run-test       |     6 -
 .../baseline.cpu.txt                          |     6 +-
 .../baseline.gpu.txt                          |     6 +-
 .../baseline.windows.cpu.txt                  |     6 +-
 .../baseline.windows.gpu.txt                  |     6 +-
 .../run-test                                  |     6 -
 .../QuickE2E/baseline.cpu.txt                 |     4 +-
 .../QuickE2E/baseline.gpu.txt                 |     4 +-
 .../QuickE2E/baseline.windows.cpu.txt         |     4 +-
 .../QuickE2E/baseline.windows.gpu.txt         |     4 +-
 .../QuickE2E/run-test                         |     6 -
 .../QuickE2E/testcases.yml                    |     8 +-
 .../SVD/baseline.cpu.txt                      |     2 +-
 .../SVD/baseline.gpu.txt                      |     2 +-
 .../SVD/baseline.windows.cpu.txt              |     2 +-
 .../SVD/baseline.windows.gpu.txt              |     2 +-
 .../ExperimentalHtkmlfReader/SVD/run-test     |     6 -
 .../SVD/testcases.yml                         |    10 +-
 .../LSTM/FullUtterance/baseline.cpu.txt       |     2 +-
 .../LSTM/FullUtterance/baseline.gpu.txt       |     2 +-
 .../FullUtterance/baseline.windows.cpu.txt    |     2 +-
 .../FullUtterance/baseline.windows.gpu.txt    |     2 +-
 .../Speech/LSTM/FullUtterance/testcases.yml   |     8 +-
 .../LSTM/Truncated-Kaldi/baseline.cpu.txt     |     2 +-
 .../LSTM/Truncated-Kaldi/baseline.gpu.txt     |     2 +-
 .../Speech/LSTM/Truncated-Kaldi/testcases.yml |     8 +-
 .../Speech/LSTM/Truncated/baseline.cpu.txt    |     2 +-
 .../Speech/LSTM/Truncated/baseline.gpu.txt    |     2 +-
 .../LSTM/Truncated/baseline.windows.cpu.txt   |     2 +-
 .../LSTM/Truncated/baseline.windows.gpu.txt   |     2 +-
 .../Speech/LSTM/Truncated/testcases.yml       |     8 +-
 .../Speech/QuickE2E/baseline.cpu.txt          |     4 +-
 .../Speech/QuickE2E/baseline.gpu.txt          |     4 +-
 .../Speech/QuickE2E/baseline.windows.cpu.txt  |     4 +-
 .../Speech/QuickE2E/baseline.windows.gpu.txt  |     4 +-
 .../Speech/QuickE2E/testcases.yml             |     8 +-
 .../EndToEndTests/Speech/SVD/baseline.cpu.txt |     2 +-
 .../EndToEndTests/Speech/SVD/baseline.gpu.txt |     2 +-
 .../Speech/SVD/baseline.windows.cpu.txt       |     2 +-
 .../Speech/SVD/baseline.windows.gpu.txt       |     2 +-
 Tests/EndToEndTests/Speech/SVD/testcases.yml  |    10 +-
 .../Speech/Simple/baseline.cpu.txt            |     4 +-
 .../Speech/Simple/baseline.gpu.txt            |     4 +-
 .../Speech/Simple/baseline.windows.cpu.txt    |     4 +-
 .../Speech/Simple/baseline.windows.gpu.txt    |     4 +-
 .../EndToEndTests/Speech/Simple/testcases.yml |     8 +-
 .../Text/SparseDSSM/baseline.cpu.txt          |    16 +-
 .../Text/SparseDSSM/baseline.gpu.txt          |    16 +-
 .../Text/SparseDSSM/baseline.windows.cpu.txt  |    16 +-
 .../Text/SparseDSSM/baseline.windows.gpu.txt  |    16 +-
 Tests/EndToEndTests/run-test-common           |     2 +-
 217 files changed, 16989 insertions(+), 16919 deletions(-)

diff --git a/Examples/Image/Miscellaneous/CIFAR-10/Output/03_ResNet_Train_AddBNEval_Test.log.160 b/Examples/Image/Miscellaneous/CIFAR-10/Output/03_ResNet_Train_AddBNEval_Test.log.160
index 2cbc9ba9e..3f57ee6c2 100644
--- a/Examples/Image/Miscellaneous/CIFAR-10/Output/03_ResNet_Train_AddBNEval_Test.log.160
+++ b/Examples/Image/Miscellaneous/CIFAR-10/Output/03_ResNet_Train_AddBNEval_Test.log.160
@@ -1,6414 +1,6414 @@
--------------------------------------------------------------------
-Build info: 
-
-		Built time: Jan 12 2016 14:46:20
-		Last modified date: Mon Jan 11 11:39:54 2016
-		CUDA_PATH: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.0
-		Build Branch: 
-		Build SHA1: 
-		Built by alexeyk on z840-01           
-		Build Path: C:\src\cntk\Source\CNTK\
--------------------------------------------------------------------
-running on z840-01 at 2016/01/14 10:36:01
-command line: 
-..\..\..\..\x64\Release\CNTK.exe configFile=03_ResNet.config 
-
->>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>
-RootDir = "."
-ConfigDir = "$RootDir$"
-DataDir = "$RootDir$"
-OutputDir = "$RootDir$/Output"
-ModelDir = "$OutputDir$/Models"
-ndlMacros=$ConfigDir$/Macros.ndl
-precision=float
-deviceId=Auto
-prefetch=true
-parallelTrain=false
-command=Train:AddBNEval:Test
-stderr=$OutputDir$/03_ResNet
-traceLevel=1
-numMBsToShowResult=200
-Proj16to32Filename = $ConfigDir$/16to32.txt
-Proj32to64Filename = $ConfigDir$/32to64.txt
-Train=[
-    action=train
-    modelPath=$ModelDir$/03_ResNet
-     NDLNetworkBuilder=[
-        networkDescription=$ConfigDir$/03_ResNet.ndl
-    ]
-    SGD=[
-        epochSize=0
-        minibatchSize=128
-        learningRatesPerMB=1.0*80:0.1*40:0.01
-        momentumPerMB=0.9
-        maxEpochs=160
-        L2RegWeight=0.0001
-        dropoutRate=0
-        ParallelTrain=[
-            parallelizationMethod=DataParallelSGD
-            distributedMBReading=true
-            parallelizationStartEpoch=1
-            DataParallelSGD=[
-                gradientBits=32
-            ]
-        ]
-    ]
-    reader=[
-        readerType=ImageReader
-        file=$DataDir$/train_map.txt
-        randomize=Auto
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType=Random
-            cropRatio=0.8
-            jitterType=UniRatio
-            interpolations=Linear
-            meanFile=$ConfigDir$/CIFAR-10_mean.xml
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-AddBNEval=[    
-    action=edit
-    CurModel=$ModelDir$/03_ResNet
-    NewModel=$ModelDir$/03_ResNet.Eval
-    editPath=$ConfigDir$/03_ResNet.mel
-]
-Test=[
-    action=test
-    modelPath=$ModelDir$/03_ResNet.Eval
-    minibatchSize=512
-     NDLNetworkBuilder=[
-        networkDescription=$ConfigDir$/03_ResNet.ndl
-    ]
-    reader=[
-        readerType=ImageReader
-        file=$DataDir$/test_map.txt
-        randomize=Auto
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType=Center
-            cropRatio=1
-            jitterType=UniRatio
-            interpolations=Linear
-            meanFile=$ConfigDir$/CIFAR-10_mean.xml
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-
-<<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<
-
->>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
-RootDir = "."
-ConfigDir = "."
-DataDir = "."
-OutputDir = "./Output"
-ModelDir = "./Output/Models"
-ndlMacros=./Macros.ndl
-precision=float
-deviceId=Auto
-prefetch=true
-parallelTrain=false
-command=Train:AddBNEval:Test
-stderr=./Output/03_ResNet
-traceLevel=1
-numMBsToShowResult=200
-Proj16to32Filename = ./16to32.txt
-Proj32to64Filename = ./32to64.txt
-Train=[
-    action=train
-    modelPath=./Output/Models/03_ResNet
-     NDLNetworkBuilder=[
-        networkDescription=./03_ResNet.ndl
-    ]
-    SGD=[
-        epochSize=0
-        minibatchSize=128
-        learningRatesPerMB=1.0*80:0.1*40:0.01
-        momentumPerMB=0.9
-        maxEpochs=160
-        L2RegWeight=0.0001
-        dropoutRate=0
-        ParallelTrain=[
-            parallelizationMethod=DataParallelSGD
-            distributedMBReading=true
-            parallelizationStartEpoch=1
-            DataParallelSGD=[
-                gradientBits=32
-            ]
-        ]
-    ]
-    reader=[
-        readerType=ImageReader
-        file=./train_map.txt
-        randomize=Auto
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType=Random
-            cropRatio=0.8
-            jitterType=UniRatio
-            interpolations=Linear
-            meanFile=./CIFAR-10_mean.xml
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-AddBNEval=[    
-    action=edit
-    CurModel=./Output/Models/03_ResNet
-    NewModel=./Output/Models/03_ResNet.Eval
-    editPath=./03_ResNet.mel
-]
-Test=[
-    action=test
-    modelPath=./Output/Models/03_ResNet.Eval
-    minibatchSize=512
-     NDLNetworkBuilder=[
-        networkDescription=./03_ResNet.ndl
-    ]
-    reader=[
-        readerType=ImageReader
-        file=./test_map.txt
-        randomize=Auto
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType=Center
-            cropRatio=1
-            jitterType=UniRatio
-            interpolations=Linear
-            meanFile=./CIFAR-10_mean.xml
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-
-<<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
-
->>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
-configparameters: 03_ResNet.config:AddBNEval=[    
-    action=edit
-    CurModel=./Output/Models/03_ResNet
-    NewModel=./Output/Models/03_ResNet.Eval
-    editPath=./03_ResNet.mel
-]
-
-configparameters: 03_ResNet.config:command=Train:AddBNEval:Test
-configparameters: 03_ResNet.config:ConfigDir=.
-configparameters: 03_ResNet.config:DataDir=.
-configparameters: 03_ResNet.config:deviceId=Auto
-configparameters: 03_ResNet.config:ModelDir=./Output/Models
-configparameters: 03_ResNet.config:ndlMacros=./Macros.ndl
-configparameters: 03_ResNet.config:numMBsToShowResult=200
-configparameters: 03_ResNet.config:OutputDir=./Output
-configparameters: 03_ResNet.config:parallelTrain=false
-configparameters: 03_ResNet.config:precision=float
-configparameters: 03_ResNet.config:prefetch=true
-configparameters: 03_ResNet.config:Proj16to32Filename=./16to32.txt
-configparameters: 03_ResNet.config:Proj32to64Filename=./32to64.txt
-configparameters: 03_ResNet.config:RootDir=.
-configparameters: 03_ResNet.config:stderr=./Output/03_ResNet
-configparameters: 03_ResNet.config:Test=[
-    action=test
-    modelPath=./Output/Models/03_ResNet.Eval
-    minibatchSize=512
-     NDLNetworkBuilder=[
-        networkDescription=./03_ResNet.ndl
-    ]
-    reader=[
-        readerType=ImageReader
-        file=./test_map.txt
-        randomize=Auto
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType=Center
-            cropRatio=1
-            jitterType=UniRatio
-            interpolations=Linear
-            meanFile=./CIFAR-10_mean.xml
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-
-configparameters: 03_ResNet.config:traceLevel=1
-configparameters: 03_ResNet.config:Train=[
-    action=train
-    modelPath=./Output/Models/03_ResNet
-     NDLNetworkBuilder=[
-        networkDescription=./03_ResNet.ndl
-    ]
-    SGD=[
-        epochSize=0
-        minibatchSize=128
-        learningRatesPerMB=1.0*80:0.1*40:0.01
-        momentumPerMB=0.9
-        maxEpochs=160
-        L2RegWeight=0.0001
-        dropoutRate=0
-        ParallelTrain=[
-            parallelizationMethod=DataParallelSGD
-            distributedMBReading=true
-            parallelizationStartEpoch=1
-            DataParallelSGD=[
-                gradientBits=32
-            ]
-        ]
-    ]
-    reader=[
-        readerType=ImageReader
-        file=./train_map.txt
-        randomize=Auto
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType=Random
-            cropRatio=0.8
-            jitterType=UniRatio
-            interpolations=Linear
-            meanFile=./CIFAR-10_mean.xml
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-
-<<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
-command: Train AddBNEval Test 
-precision = float
-CNTKModelPath: ./Output/Models/03_ResNet
-CNTKCommandTrainInfo: Train : 160
-CNTKCommandTrainInfo: CNTKNoMoreCommands_Total : 160
-CNTKCommandTrainBegin: Train
-LockDevice: Locked GPU 0 to test availability.
-LockDevice: Unlocked GPU 0 after testing.
-LockDevice: Locked GPU 2 to test availability.
-LockDevice: Unlocked GPU 2 after testing.
-LockDevice: Locked GPU 1 to test availability.
-LockDevice: Unlocked GPU 1 after testing.
-LockDevice: Locked GPU 0 for exclusive use.
-NDLBuilder Using GPU 0
-Microsoft::MSR::CNTK::GPUMatrix<ElemType>::SetGaussianRandomValue (GPU): creating curand object with seed 1, sizeof(ElemType)==4
-
-Post-processing network...
-
-3 roots:
-	Err = ErrorPrediction
-	OutputNodes.z = Plus
-	CE = CrossEntropyWithSoftmax
-FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
-FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
-FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
-
-
-Validating for node Err. 187 nodes to process in pass 1.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node Err. 78 nodes to process in pass 2.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node Err, final verification.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-108 out of 187 nodes do not share the minibatch layout with the input data.
-
-
-Validating for node OutputNodes.z. 185 nodes to process in pass 1.
-
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-
-Validating for node OutputNodes.z. 77 nodes to process in pass 2.
-
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-
-Validating for node OutputNodes.z, final verification.
-
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-
-107 out of 185 nodes do not share the minibatch layout with the input data.
-
-
-Validating for node CE. 187 nodes to process in pass 1.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node CE. 78 nodes to process in pass 2.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node CE, final verification.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-108 out of 187 nodes do not share the minibatch layout with the input data.
-
-Post-processing network complete.
-
-SGD using GPU 0.
-
-Training criterion node(s):
-	CE = CrossEntropyWithSoftmax
-
-Evaluation criterion node(s):
-	Err = ErrorPrediction
-
-
-Allocating matrices for forward and/or backward propagation.
-No PreCompute nodes found, skipping PreCompute step
-Set Max Temp Mem Size For Convolution Nodes to 0 samples.
-Starting Epoch 1: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
-#NLop10#
-Tensor Op: Op 10: 32 x 32 x 16 x 128 x 1 -> 32 x 32 x 16 x 128 x 1
-24 procs  32 warps  2147483647 65535 65535 max grid  on  GeForce GTX TITAN X
-3 procs  32 warps  2147483647 65535 65535 max grid  on  Quadro K620
-14 procs  32 warps  2147483647 65535 65535 max grid  on  GeForce GTX TITAN
-Tensor Op: Op 15: 32 x 32 x 16 x 128 x 1 op 32 x 32 x 16 x 128 x 1 -> 32 x 32 x 16 x 128 x 1
- Epoch[ 1 of 160]-Minibatch[   1- 200]: SamplesSeen = 25600; TrainLossPerSample =  1.78511215; EvalErr[0]PerSample = 0.67218750; TotalTime = 15.4090s; SamplesPerSecond = 1661.4
-Finished Epoch[ 1 of 160]: [Training Set] TrainLossPerSample = 1.6011882; EvalErrPerSample = 0.59435999; AvgLearningRatePerSample = 0.0078125; EpochTime=26.129
-Starting Epoch 2: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 2 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  1.18311264; EvalErr[0]PerSample = 0.42843750; TotalTime = 11.4034s; SamplesPerSecond = 2245.0
-Finished Epoch[ 2 of 160]: [Training Set] TrainLossPerSample = 1.1033459; EvalErrPerSample = 0.3969; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1287
-Starting Epoch 3: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 3 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.90734154; EvalErr[0]PerSample = 0.32207031; TotalTime = 11.3750s; SamplesPerSecond = 2250.5
-Finished Epoch[ 3 of 160]: [Training Set] TrainLossPerSample = 0.86615896; EvalErrPerSample = 0.30719998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.0657
-Starting Epoch 4: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 4 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.77141624; EvalErr[0]PerSample = 0.27066406; TotalTime = 11.4567s; SamplesPerSecond = 2234.5
-Finished Epoch[ 4 of 160]: [Training Set] TrainLossPerSample = 0.75146842; EvalErrPerSample = 0.26264; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2146
-Starting Epoch 5: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 5 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.68756187; EvalErr[0]PerSample = 0.23796875; TotalTime = 11.4042s; SamplesPerSecond = 2244.8
-Finished Epoch[ 5 of 160]: [Training Set] TrainLossPerSample = 0.67170274; EvalErrPerSample = 0.23289999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1571
-Starting Epoch 6: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 6 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.61656025; EvalErr[0]PerSample = 0.21398438; TotalTime = 11.4800s; SamplesPerSecond = 2230.0
-Finished Epoch[ 6 of 160]: [Training Set] TrainLossPerSample = 0.612014; EvalErrPerSample = 0.21263999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2553
-Starting Epoch 7: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 7 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.58124241; EvalErr[0]PerSample = 0.20335938; TotalTime = 11.4640s; SamplesPerSecond = 2233.1
-Finished Epoch[ 7 of 160]: [Training Set] TrainLossPerSample = 0.56962705; EvalErrPerSample = 0.19909999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2453
-Starting Epoch 8: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 8 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.54073845; EvalErr[0]PerSample = 0.18796875; TotalTime = 11.5047s; SamplesPerSecond = 2225.2
-Finished Epoch[ 8 of 160]: [Training Set] TrainLossPerSample = 0.53796959; EvalErrPerSample = 0.18616; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3345
-Starting Epoch 9: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 9 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.50834869; EvalErr[0]PerSample = 0.17632813; TotalTime = 11.5167s; SamplesPerSecond = 2222.9
-Finished Epoch[ 9 of 160]: [Training Set] TrainLossPerSample = 0.51177925; EvalErrPerSample = 0.17704; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3221
-Starting Epoch 10: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[10 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.48615803; EvalErr[0]PerSample = 0.16917969; TotalTime = 11.4815s; SamplesPerSecond = 2229.7
-Finished Epoch[10 of 160]: [Training Set] TrainLossPerSample = 0.49343586; EvalErrPerSample = 0.17158; AvgLearningRatePerSample = 0.0078125; EpochTime=22.286
-Starting Epoch 11: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[11 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.46868595; EvalErr[0]PerSample = 0.16312500; TotalTime = 11.4115s; SamplesPerSecond = 2243.4
-Finished Epoch[11 of 160]: [Training Set] TrainLossPerSample = 0.46725979; EvalErrPerSample = 0.16214; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2195
-Starting Epoch 12: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[12 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.45617180; EvalErr[0]PerSample = 0.15531250; TotalTime = 11.4316s; SamplesPerSecond = 2239.4
-Finished Epoch[12 of 160]: [Training Set] TrainLossPerSample = 0.44944596; EvalErrPerSample = 0.15497999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2204
-Starting Epoch 13: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[13 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.43442238; EvalErr[0]PerSample = 0.15031250; TotalTime = 11.5970s; SamplesPerSecond = 2207.5
-Finished Epoch[13 of 160]: [Training Set] TrainLossPerSample = 0.43854889; EvalErrPerSample = 0.15235999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4304
-Starting Epoch 14: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[14 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.41474213; EvalErr[0]PerSample = 0.14265625; TotalTime = 11.5551s; SamplesPerSecond = 2215.5
-Finished Epoch[14 of 160]: [Training Set] TrainLossPerSample = 0.42394838; EvalErrPerSample = 0.14659999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3929
-Starting Epoch 15: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[15 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.40827648; EvalErr[0]PerSample = 0.14167969; TotalTime = 11.5593s; SamplesPerSecond = 2214.7
-Finished Epoch[15 of 160]: [Training Set] TrainLossPerSample = 0.40738714; EvalErrPerSample = 0.14126; AvgLearningRatePerSample = 0.0078125; EpochTime=22.404
-Starting Epoch 16: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[16 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.39572418; EvalErr[0]PerSample = 0.13765625; TotalTime = 11.6490s; SamplesPerSecond = 2197.6
-Finished Epoch[16 of 160]: [Training Set] TrainLossPerSample = 0.4007735; EvalErrPerSample = 0.13798; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4156
-Starting Epoch 17: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[17 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.38931370; EvalErr[0]PerSample = 0.13757813; TotalTime = 11.4136s; SamplesPerSecond = 2242.9
-Finished Epoch[17 of 160]: [Training Set] TrainLossPerSample = 0.39249194; EvalErrPerSample = 0.13716; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1748
-Starting Epoch 18: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[18 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.37350758; EvalErr[0]PerSample = 0.12980469; TotalTime = 11.5431s; SamplesPerSecond = 2217.8
-Finished Epoch[18 of 160]: [Training Set] TrainLossPerSample = 0.37985951; EvalErrPerSample = 0.13271999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3177
-Starting Epoch 19: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[19 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.37046688; EvalErr[0]PerSample = 0.12742187; TotalTime = 11.4904s; SamplesPerSecond = 2228.0
-Finished Epoch[19 of 160]: [Training Set] TrainLossPerSample = 0.37240577; EvalErrPerSample = 0.12842; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2594
-Starting Epoch 20: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[20 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.36126266; EvalErr[0]PerSample = 0.12656250; TotalTime = 11.5121s; SamplesPerSecond = 2223.7
-Finished Epoch[20 of 160]: [Training Set] TrainLossPerSample = 0.36667159; EvalErrPerSample = 0.12833999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2756
-Starting Epoch 21: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[21 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.36016537; EvalErr[0]PerSample = 0.12242188; TotalTime = 11.5126s; SamplesPerSecond = 2223.7
-Finished Epoch[21 of 160]: [Training Set] TrainLossPerSample = 0.35836998; EvalErrPerSample = 0.12233999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2811
-Starting Epoch 22: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[22 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.34188179; EvalErr[0]PerSample = 0.11910156; TotalTime = 11.5167s; SamplesPerSecond = 2222.9
-Finished Epoch[22 of 160]: [Training Set] TrainLossPerSample = 0.34598714; EvalErrPerSample = 0.12013999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3214
-Starting Epoch 23: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[23 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.34518520; EvalErr[0]PerSample = 0.11894531; TotalTime = 11.5177s; SamplesPerSecond = 2222.7
-Finished Epoch[23 of 160]: [Training Set] TrainLossPerSample = 0.34407225; EvalErrPerSample = 0.11892; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3061
-Starting Epoch 24: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[24 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.33347424; EvalErr[0]PerSample = 0.11660156; TotalTime = 11.5769s; SamplesPerSecond = 2211.3
-Finished Epoch[24 of 160]: [Training Set] TrainLossPerSample = 0.34192094; EvalErrPerSample = 0.11848; AvgLearningRatePerSample = 0.0078125; EpochTime=22.368
-Starting Epoch 25: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[25 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.33844280; EvalErr[0]PerSample = 0.11648438; TotalTime = 11.5529s; SamplesPerSecond = 2215.9
-Finished Epoch[25 of 160]: [Training Set] TrainLossPerSample = 0.3405562; EvalErrPerSample = 0.11768; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3158
-Starting Epoch 26: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[26 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.32573421; EvalErr[0]PerSample = 0.11328125; TotalTime = 11.4885s; SamplesPerSecond = 2228.3
-Finished Epoch[26 of 160]: [Training Set] TrainLossPerSample = 0.3283667; EvalErrPerSample = 0.11268; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3241
-Starting Epoch 27: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[27 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.32139095; EvalErr[0]PerSample = 0.11300781; TotalTime = 11.5773s; SamplesPerSecond = 2211.2
-Finished Epoch[27 of 160]: [Training Set] TrainLossPerSample = 0.32817245; EvalErrPerSample = 0.11406; AvgLearningRatePerSample = 0.0078125; EpochTime=22.345
-Starting Epoch 28: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[28 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.31476902; EvalErr[0]PerSample = 0.11074219; TotalTime = 11.4481s; SamplesPerSecond = 2236.2
-Finished Epoch[28 of 160]: [Training Set] TrainLossPerSample = 0.32170638; EvalErrPerSample = 0.1133; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2168
-Starting Epoch 29: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[29 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30912992; EvalErr[0]PerSample = 0.10859375; TotalTime = 11.4067s; SamplesPerSecond = 2244.3
-Finished Epoch[29 of 160]: [Training Set] TrainLossPerSample = 0.31786552; EvalErrPerSample = 0.11086; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1891
-Starting Epoch 30: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[30 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30547693; EvalErr[0]PerSample = 0.10496094; TotalTime = 11.5656s; SamplesPerSecond = 2213.5
-Finished Epoch[30 of 160]: [Training Set] TrainLossPerSample = 0.31253323; EvalErrPerSample = 0.1078; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3465
-Starting Epoch 31: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[31 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30724499; EvalErr[0]PerSample = 0.10769531; TotalTime = 11.5380s; SamplesPerSecond = 2218.8
-Finished Epoch[31 of 160]: [Training Set] TrainLossPerSample = 0.31017771; EvalErrPerSample = 0.10802; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3988
-Starting Epoch 32: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[32 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30443895; EvalErr[0]PerSample = 0.10660156; TotalTime = 11.5474s; SamplesPerSecond = 2216.9
-Finished Epoch[32 of 160]: [Training Set] TrainLossPerSample = 0.31054172; EvalErrPerSample = 0.10764; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3142
-Starting Epoch 33: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[33 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.29248705; EvalErr[0]PerSample = 0.10175781; TotalTime = 11.4051s; SamplesPerSecond = 2244.6
-Finished Epoch[33 of 160]: [Training Set] TrainLossPerSample = 0.30296284; EvalErrPerSample = 0.10422; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1968
-Starting Epoch 34: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[34 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30263618; EvalErr[0]PerSample = 0.10550781; TotalTime = 11.5888s; SamplesPerSecond = 2209.0
-Finished Epoch[34 of 160]: [Training Set] TrainLossPerSample = 0.30190024; EvalErrPerSample = 0.10473999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4491
-Starting Epoch 35: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[35 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.29659115; EvalErr[0]PerSample = 0.10171875; TotalTime = 11.6140s; SamplesPerSecond = 2204.2
-Finished Epoch[35 of 160]: [Training Set] TrainLossPerSample = 0.29864648; EvalErrPerSample = 0.10306; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4411
-Starting Epoch 36: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[36 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28713980; EvalErr[0]PerSample = 0.10027344; TotalTime = 11.5617s; SamplesPerSecond = 2214.2
-Finished Epoch[36 of 160]: [Training Set] TrainLossPerSample = 0.29245624; EvalErrPerSample = 0.10234; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3455
-Starting Epoch 37: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[37 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28737648; EvalErr[0]PerSample = 0.09847656; TotalTime = 11.4851s; SamplesPerSecond = 2229.0
-Finished Epoch[37 of 160]: [Training Set] TrainLossPerSample = 0.29400381; EvalErrPerSample = 0.10157999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3035
-Starting Epoch 38: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[38 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28604475; EvalErr[0]PerSample = 0.09699219; TotalTime = 11.6237s; SamplesPerSecond = 2202.4
-Finished Epoch[38 of 160]: [Training Set] TrainLossPerSample = 0.29055047; EvalErrPerSample = 0.1006; AvgLearningRatePerSample = 0.0078125; EpochTime=22.41
-Starting Epoch 39: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[39 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28677174; EvalErr[0]PerSample = 0.09929688; TotalTime = 11.6037s; SamplesPerSecond = 2206.2
-Finished Epoch[39 of 160]: [Training Set] TrainLossPerSample = 0.28853956; EvalErrPerSample = 0.101; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4156
-Starting Epoch 40: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[40 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27731121; EvalErr[0]PerSample = 0.09699219; TotalTime = 11.5688s; SamplesPerSecond = 2212.9
-Finished Epoch[40 of 160]: [Training Set] TrainLossPerSample = 0.28800672; EvalErrPerSample = 0.099959999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3445
-Starting Epoch 41: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[41 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28877285; EvalErr[0]PerSample = 0.10156250; TotalTime = 11.5167s; SamplesPerSecond = 2222.9
-Finished Epoch[41 of 160]: [Training Set] TrainLossPerSample = 0.28422225; EvalErrPerSample = 0.098719999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.355
-Starting Epoch 42: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[42 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27758221; EvalErr[0]PerSample = 0.09710937; TotalTime = 11.5935s; SamplesPerSecond = 2208.1
-Finished Epoch[42 of 160]: [Training Set] TrainLossPerSample = 0.28059965; EvalErrPerSample = 0.097819999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.7852
-Starting Epoch 43: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[43 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27680891; EvalErr[0]PerSample = 0.09644531; TotalTime = 11.6198s; SamplesPerSecond = 2203.1
-Finished Epoch[43 of 160]: [Training Set] TrainLossPerSample = 0.28217828; EvalErrPerSample = 0.098200001; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4529
-Starting Epoch 44: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[44 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27107407; EvalErr[0]PerSample = 0.09421875; TotalTime = 11.5929s; SamplesPerSecond = 2208.2
-Finished Epoch[44 of 160]: [Training Set] TrainLossPerSample = 0.27688959; EvalErrPerSample = 0.096079998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3608
-Starting Epoch 45: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[45 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27329311; EvalErr[0]PerSample = 0.09585937; TotalTime = 11.4535s; SamplesPerSecond = 2235.1
-Finished Epoch[45 of 160]: [Training Set] TrainLossPerSample = 0.27860796; EvalErrPerSample = 0.097539999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2187
-Starting Epoch 46: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[46 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.26291889; EvalErr[0]PerSample = 0.09253906; TotalTime = 11.4315s; SamplesPerSecond = 2239.4
-Finished Epoch[46 of 160]: [Training Set] TrainLossPerSample = 0.27177998; EvalErrPerSample = 0.095179997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1909
-Starting Epoch 47: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[47 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27016382; EvalErr[0]PerSample = 0.09503906; TotalTime = 11.4016s; SamplesPerSecond = 2245.3
-Finished Epoch[47 of 160]: [Training Set] TrainLossPerSample = 0.27474955; EvalErrPerSample = 0.096859999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1576
-Starting Epoch 48: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[48 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25722231; EvalErr[0]PerSample = 0.08960938; TotalTime = 11.4404s; SamplesPerSecond = 2237.7
-Finished Epoch[48 of 160]: [Training Set] TrainLossPerSample = 0.26707178; EvalErrPerSample = 0.09296; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2165
-Starting Epoch 49: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[49 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25658318; EvalErr[0]PerSample = 0.08960938; TotalTime = 11.5004s; SamplesPerSecond = 2226.0
-Finished Epoch[49 of 160]: [Training Set] TrainLossPerSample = 0.26653028; EvalErrPerSample = 0.093059994; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2617
-Starting Epoch 50: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[50 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.26309399; EvalErr[0]PerSample = 0.09062500; TotalTime = 11.4840s; SamplesPerSecond = 2229.2
-Finished Epoch[50 of 160]: [Training Set] TrainLossPerSample = 0.27242121; EvalErrPerSample = 0.094359994; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4308
-Starting Epoch 51: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[51 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25116308; EvalErr[0]PerSample = 0.08757813; TotalTime = 11.5958s; SamplesPerSecond = 2207.7
-Finished Epoch[51 of 160]: [Training Set] TrainLossPerSample = 0.26714918; EvalErrPerSample = 0.093719997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3675
-Starting Epoch 52: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[52 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25991890; EvalErr[0]PerSample = 0.09207031; TotalTime = 11.4533s; SamplesPerSecond = 2235.2
-Finished Epoch[52 of 160]: [Training Set] TrainLossPerSample = 0.2633768; EvalErrPerSample = 0.09268; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3111
-Starting Epoch 53: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[53 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25488058; EvalErr[0]PerSample = 0.08957031; TotalTime = 11.5890s; SamplesPerSecond = 2209.0
-Finished Epoch[53 of 160]: [Training Set] TrainLossPerSample = 0.26292121; EvalErrPerSample = 0.092160001; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3534
-Starting Epoch 54: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[54 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24893934; EvalErr[0]PerSample = 0.08859375; TotalTime = 11.5135s; SamplesPerSecond = 2223.5
-Finished Epoch[54 of 160]: [Training Set] TrainLossPerSample = 0.25717214; EvalErrPerSample = 0.090659998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4264
-Starting Epoch 55: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[55 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25861763; EvalErr[0]PerSample = 0.09019531; TotalTime = 11.6295s; SamplesPerSecond = 2201.3
-Finished Epoch[55 of 160]: [Training Set] TrainLossPerSample = 0.26373541; EvalErrPerSample = 0.091260001; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3938
-Starting Epoch 56: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[56 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25139614; EvalErr[0]PerSample = 0.08546875; TotalTime = 11.5303s; SamplesPerSecond = 2220.2
-Finished Epoch[56 of 160]: [Training Set] TrainLossPerSample = 0.25967881; EvalErrPerSample = 0.089499995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2997
-Starting Epoch 57: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[57 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24860460; EvalErr[0]PerSample = 0.08734375; TotalTime = 11.5494s; SamplesPerSecond = 2216.6
-Finished Epoch[57 of 160]: [Training Set] TrainLossPerSample = 0.25601548; EvalErrPerSample = 0.090059996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3412
-Starting Epoch 58: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[58 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24636944; EvalErr[0]PerSample = 0.08496094; TotalTime = 11.5079s; SamplesPerSecond = 2224.6
-Finished Epoch[58 of 160]: [Training Set] TrainLossPerSample = 0.25675642; EvalErrPerSample = 0.088599995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3309
-Starting Epoch 59: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[59 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25354239; EvalErr[0]PerSample = 0.09003906; TotalTime = 11.5633s; SamplesPerSecond = 2213.9
-Finished Epoch[59 of 160]: [Training Set] TrainLossPerSample = 0.25516802; EvalErrPerSample = 0.090879999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3313
-Starting Epoch 60: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[60 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24986008; EvalErr[0]PerSample = 0.08527344; TotalTime = 11.4490s; SamplesPerSecond = 2236.0
-Finished Epoch[60 of 160]: [Training Set] TrainLossPerSample = 0.25723642; EvalErrPerSample = 0.088979997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.242
-Starting Epoch 61: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[61 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24366058; EvalErr[0]PerSample = 0.08671875; TotalTime = 11.5741s; SamplesPerSecond = 2211.8
-Finished Epoch[61 of 160]: [Training Set] TrainLossPerSample = 0.25719753; EvalErrPerSample = 0.091239996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3397
-Starting Epoch 62: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[62 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24259378; EvalErr[0]PerSample = 0.08492187; TotalTime = 11.5269s; SamplesPerSecond = 2220.9
-Finished Epoch[62 of 160]: [Training Set] TrainLossPerSample = 0.2515817; EvalErrPerSample = 0.088679999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3014
-Starting Epoch 63: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[63 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24029449; EvalErr[0]PerSample = 0.08460938; TotalTime = 11.5659s; SamplesPerSecond = 2213.4
-Finished Epoch[63 of 160]: [Training Set] TrainLossPerSample = 0.25265002; EvalErrPerSample = 0.088959999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3815
-Starting Epoch 64: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[64 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23263992; EvalErr[0]PerSample = 0.08144531; TotalTime = 11.5963s; SamplesPerSecond = 2207.6
-Finished Epoch[64 of 160]: [Training Set] TrainLossPerSample = 0.24182168; EvalErrPerSample = 0.084859997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4068
-Starting Epoch 65: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[65 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24372847; EvalErr[0]PerSample = 0.08410156; TotalTime = 11.5519s; SamplesPerSecond = 2216.1
-Finished Epoch[65 of 160]: [Training Set] TrainLossPerSample = 0.25128376; EvalErrPerSample = 0.087699994; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3194
-Starting Epoch 66: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[66 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23706863; EvalErr[0]PerSample = 0.08332031; TotalTime = 11.5141s; SamplesPerSecond = 2223.4
-Finished Epoch[66 of 160]: [Training Set] TrainLossPerSample = 0.2469063; EvalErrPerSample = 0.085879996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2777
-Starting Epoch 67: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[67 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23676601; EvalErr[0]PerSample = 0.08445313; TotalTime = 11.4886s; SamplesPerSecond = 2228.3
-Finished Epoch[67 of 160]: [Training Set] TrainLossPerSample = 0.2486082; EvalErrPerSample = 0.087859996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2538
-Starting Epoch 68: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[68 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23655962; EvalErr[0]PerSample = 0.08148437; TotalTime = 11.5050s; SamplesPerSecond = 2225.1
-Finished Epoch[68 of 160]: [Training Set] TrainLossPerSample = 0.24334124; EvalErrPerSample = 0.084699996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2913
-Starting Epoch 69: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[69 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23345692; EvalErr[0]PerSample = 0.08085937; TotalTime = 11.5400s; SamplesPerSecond = 2218.4
-Finished Epoch[69 of 160]: [Training Set] TrainLossPerSample = 0.241666; EvalErrPerSample = 0.083919995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3544
-Starting Epoch 70: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[70 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23493608; EvalErr[0]PerSample = 0.08160156; TotalTime = 11.5393s; SamplesPerSecond = 2218.5
-Finished Epoch[70 of 160]: [Training Set] TrainLossPerSample = 0.24537624; EvalErrPerSample = 0.084799998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3067
-Starting Epoch 71: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[71 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23666447; EvalErr[0]PerSample = 0.08300781; TotalTime = 11.5368s; SamplesPerSecond = 2219.0
-Finished Epoch[71 of 160]: [Training Set] TrainLossPerSample = 0.24771863; EvalErrPerSample = 0.086999997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3618
-Starting Epoch 72: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[72 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24186720; EvalErr[0]PerSample = 0.08511719; TotalTime = 11.5608s; SamplesPerSecond = 2214.4
-Finished Epoch[72 of 160]: [Training Set] TrainLossPerSample = 0.2442316; EvalErrPerSample = 0.085159995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3384
-Starting Epoch 73: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[73 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23745417; EvalErr[0]PerSample = 0.08406250; TotalTime = 11.5428s; SamplesPerSecond = 2217.8
-Finished Epoch[73 of 160]: [Training Set] TrainLossPerSample = 0.24105896; EvalErrPerSample = 0.084639996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3139
-Starting Epoch 74: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[74 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25070684; EvalErr[0]PerSample = 0.08886719; TotalTime = 11.4398s; SamplesPerSecond = 2237.8
-Finished Epoch[74 of 160]: [Training Set] TrainLossPerSample = 0.24804311; EvalErrPerSample = 0.087299995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2443
-Starting Epoch 75: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[75 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23875351; EvalErr[0]PerSample = 0.08117188; TotalTime = 11.5714s; SamplesPerSecond = 2212.3
-Finished Epoch[75 of 160]: [Training Set] TrainLossPerSample = 0.24082842; EvalErrPerSample = 0.083080001; AvgLearningRatePerSample = 0.0078125; EpochTime=22.366
-Starting Epoch 76: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[76 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23140110; EvalErr[0]PerSample = 0.08203125; TotalTime = 11.5559s; SamplesPerSecond = 2215.3
-Finished Epoch[76 of 160]: [Training Set] TrainLossPerSample = 0.24258973; EvalErrPerSample = 0.084919997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3202
-Starting Epoch 77: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[77 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23099312; EvalErr[0]PerSample = 0.08105469; TotalTime = 11.5710s; SamplesPerSecond = 2212.4
-Finished Epoch[77 of 160]: [Training Set] TrainLossPerSample = 0.24171358; EvalErrPerSample = 0.084399998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.416
-Starting Epoch 78: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[78 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.22422340; EvalErr[0]PerSample = 0.07945312; TotalTime = 11.5828s; SamplesPerSecond = 2210.2
-Finished Epoch[78 of 160]: [Training Set] TrainLossPerSample = 0.23471047; EvalErrPerSample = 0.083219998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3526
-Starting Epoch 79: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[79 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23049221; EvalErr[0]PerSample = 0.08292969; TotalTime = 11.5700s; SamplesPerSecond = 2212.6
-Finished Epoch[79 of 160]: [Training Set] TrainLossPerSample = 0.237709; EvalErrPerSample = 0.083939999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3438
-Starting Epoch 80: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[80 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23512028; EvalErr[0]PerSample = 0.08218750; TotalTime = 11.4967s; SamplesPerSecond = 2226.7
-Finished Epoch[80 of 160]: [Training Set] TrainLossPerSample = 0.23986502; EvalErrPerSample = 0.084139995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.32
-Starting Epoch 81: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[81 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18993895; EvalErr[0]PerSample = 0.06554688; TotalTime = 11.5483s; SamplesPerSecond = 2216.8
-Finished Epoch[81 of 160]: [Training Set] TrainLossPerSample = 0.16793214; EvalErrPerSample = 0.057559997; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3134
-Starting Epoch 82: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[82 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.13101171; EvalErr[0]PerSample = 0.04542969; TotalTime = 11.5430s; SamplesPerSecond = 2217.8
-Finished Epoch[82 of 160]: [Training Set] TrainLossPerSample = 0.12739825; EvalErrPerSample = 0.04394; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3113
-Starting Epoch 83: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[83 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.11582967; EvalErr[0]PerSample = 0.03949219; TotalTime = 11.5821s; SamplesPerSecond = 2210.3
-Finished Epoch[83 of 160]: [Training Set] TrainLossPerSample = 0.11576424; EvalErrPerSample = 0.03926; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.4228
-Starting Epoch 84: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[84 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.10656901; EvalErr[0]PerSample = 0.03722656; TotalTime = 11.5453s; SamplesPerSecond = 2217.4
-Finished Epoch[84 of 160]: [Training Set] TrainLossPerSample = 0.106635; EvalErrPerSample = 0.03644; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.5034
-Starting Epoch 85: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[85 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.09979140; EvalErr[0]PerSample = 0.03359375; TotalTime = 11.4506s; SamplesPerSecond = 2235.7
-Finished Epoch[85 of 160]: [Training Set] TrainLossPerSample = 0.10094135; EvalErrPerSample = 0.033980001; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2631
-Starting Epoch 86: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[86 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.09689341; EvalErr[0]PerSample = 0.03265625; TotalTime = 11.5657s; SamplesPerSecond = 2213.4
-Finished Epoch[86 of 160]: [Training Set] TrainLossPerSample = 0.096568428; EvalErrPerSample = 0.033059999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3286
-Starting Epoch 87: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[87 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.09459194; EvalErr[0]PerSample = 0.03140625; TotalTime = 11.4668s; SamplesPerSecond = 2232.5
-Finished Epoch[87 of 160]: [Training Set] TrainLossPerSample = 0.09401314; EvalErrPerSample = 0.031119999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2404
-Starting Epoch 88: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[88 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.09019656; EvalErr[0]PerSample = 0.02929688; TotalTime = 11.4755s; SamplesPerSecond = 2230.8
-Finished Epoch[88 of 160]: [Training Set] TrainLossPerSample = 0.090820439; EvalErrPerSample = 0.030199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2533
-Starting Epoch 89: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[89 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.08303898; EvalErr[0]PerSample = 0.02785156; TotalTime = 11.5285s; SamplesPerSecond = 2220.6
-Finished Epoch[89 of 160]: [Training Set] TrainLossPerSample = 0.086739741; EvalErrPerSample = 0.02956; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.307
-Starting Epoch 90: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[90 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.08771500; EvalErr[0]PerSample = 0.02984375; TotalTime = 11.5321s; SamplesPerSecond = 2219.9
-Finished Epoch[90 of 160]: [Training Set] TrainLossPerSample = 0.087390222; EvalErrPerSample = 0.03018; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3574
-Starting Epoch 91: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[91 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.08021155; EvalErr[0]PerSample = 0.02699219; TotalTime = 11.5502s; SamplesPerSecond = 2216.4
-Finished Epoch[91 of 160]: [Training Set] TrainLossPerSample = 0.080674298; EvalErrPerSample = 0.027319999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3191
-Starting Epoch 92: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[92 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.08322118; EvalErr[0]PerSample = 0.02910156; TotalTime = 11.5210s; SamplesPerSecond = 2222.0
-Finished Epoch[92 of 160]: [Training Set] TrainLossPerSample = 0.08106219; EvalErrPerSample = 0.028339999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2938
-Starting Epoch 93: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[93 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.08073726; EvalErr[0]PerSample = 0.02789062; TotalTime = 11.3990s; SamplesPerSecond = 2245.8
-Finished Epoch[93 of 160]: [Training Set] TrainLossPerSample = 0.079686686; EvalErrPerSample = 0.027419999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2285
-Starting Epoch 94: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[94 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07666363; EvalErr[0]PerSample = 0.02695313; TotalTime = 11.5560s; SamplesPerSecond = 2215.3
-Finished Epoch[94 of 160]: [Training Set] TrainLossPerSample = 0.077968039; EvalErrPerSample = 0.027099999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3268
-Starting Epoch 95: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[95 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07550663; EvalErr[0]PerSample = 0.02609375; TotalTime = 11.4795s; SamplesPerSecond = 2230.1
-Finished Epoch[95 of 160]: [Training Set] TrainLossPerSample = 0.076805077; EvalErrPerSample = 0.02616; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2683
-Starting Epoch 96: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[96 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07483893; EvalErr[0]PerSample = 0.02652344; TotalTime = 11.4940s; SamplesPerSecond = 2227.3
-Finished Epoch[96 of 160]: [Training Set] TrainLossPerSample = 0.073900625; EvalErrPerSample = 0.025839999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2583
-Starting Epoch 97: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[97 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07033884; EvalErr[0]PerSample = 0.02421875; TotalTime = 11.4972s; SamplesPerSecond = 2226.6
-Finished Epoch[97 of 160]: [Training Set] TrainLossPerSample = 0.072241709; EvalErrPerSample = 0.025239998; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3373
-Starting Epoch 98: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[98 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07043020; EvalErr[0]PerSample = 0.02421875; TotalTime = 11.5606s; SamplesPerSecond = 2214.4
-Finished Epoch[98 of 160]: [Training Set] TrainLossPerSample = 0.070759542; EvalErrPerSample = 0.024279999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.328
-Starting Epoch 99: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[99 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07046758; EvalErr[0]PerSample = 0.02503906; TotalTime = 11.5808s; SamplesPerSecond = 2210.5
-Finished Epoch[99 of 160]: [Training Set] TrainLossPerSample = 0.069847755; EvalErrPerSample = 0.024219999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3476
-Starting Epoch 100: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[100 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.06165749; EvalErr[0]PerSample = 0.02144531; TotalTime = 11.5168s; SamplesPerSecond = 2222.8
-Finished Epoch[100 of 160]: [Training Set] TrainLossPerSample = 0.066228345; EvalErrPerSample = 0.02266; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2882
-Starting Epoch 101: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[101 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.06483539; EvalErr[0]PerSample = 0.02242188; TotalTime = 11.5478s; SamplesPerSecond = 2216.9
-Finished Epoch[101 of 160]: [Training Set] TrainLossPerSample = 0.066535138; EvalErrPerSample = 0.022839999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.382
-Starting Epoch 102: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[102 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.06374743; EvalErr[0]PerSample = 0.02113281; TotalTime = 11.7412s; SamplesPerSecond = 2180.4
-Finished Epoch[102 of 160]: [Training Set] TrainLossPerSample = 0.06468109; EvalErrPerSample = 0.021979999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.5325
-Starting Epoch 103: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[103 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.06524000; EvalErr[0]PerSample = 0.02261719; TotalTime = 11.5928s; SamplesPerSecond = 2208.3
-Finished Epoch[103 of 160]: [Training Set] TrainLossPerSample = 0.065452188; EvalErrPerSample = 0.021879999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3876
-Starting Epoch 104: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[104 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.06336018; EvalErr[0]PerSample = 0.02117188; TotalTime = 11.5741s; SamplesPerSecond = 2211.8
-Finished Epoch[104 of 160]: [Training Set] TrainLossPerSample = 0.064649269; EvalErrPerSample = 0.021819999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3409
-Starting Epoch 105: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[105 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05928923; EvalErr[0]PerSample = 0.02039062; TotalTime = 11.4805s; SamplesPerSecond = 2229.9
-Finished Epoch[105 of 160]: [Training Set] TrainLossPerSample = 0.062256567; EvalErrPerSample = 0.021199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2864
-Starting Epoch 106: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[106 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05956097; EvalErr[0]PerSample = 0.02070312; TotalTime = 11.5877s; SamplesPerSecond = 2209.2
-Finished Epoch[106 of 160]: [Training Set] TrainLossPerSample = 0.059500545; EvalErrPerSample = 0.02076; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3573
-Starting Epoch 107: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[107 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05682479; EvalErr[0]PerSample = 0.01898438; TotalTime = 11.4687s; SamplesPerSecond = 2232.2
-Finished Epoch[107 of 160]: [Training Set] TrainLossPerSample = 0.05800005; EvalErrPerSample = 0.019579999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3035
-Starting Epoch 108: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[108 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05657757; EvalErr[0]PerSample = 0.01890625; TotalTime = 11.5761s; SamplesPerSecond = 2211.4
-Finished Epoch[108 of 160]: [Training Set] TrainLossPerSample = 0.059094436; EvalErrPerSample = 0.02004; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3386
-Starting Epoch 109: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[109 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05917750; EvalErr[0]PerSample = 0.01945313; TotalTime = 11.5089s; SamplesPerSecond = 2224.4
-Finished Epoch[109 of 160]: [Training Set] TrainLossPerSample = 0.057597388; EvalErrPerSample = 0.01898; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2734
-Starting Epoch 110: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[110 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05674717; EvalErr[0]PerSample = 0.01968750; TotalTime = 11.5447s; SamplesPerSecond = 2217.5
-Finished Epoch[110 of 160]: [Training Set] TrainLossPerSample = 0.05731637; EvalErrPerSample = 0.019819999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3039
-Starting Epoch 111: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[111 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05444533; EvalErr[0]PerSample = 0.01945313; TotalTime = 11.4688s; SamplesPerSecond = 2232.1
-Finished Epoch[111 of 160]: [Training Set] TrainLossPerSample = 0.055398725; EvalErrPerSample = 0.01956; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.244
-Starting Epoch 112: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[112 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05179993; EvalErr[0]PerSample = 0.01710938; TotalTime = 11.5138s; SamplesPerSecond = 2223.4
-Finished Epoch[112 of 160]: [Training Set] TrainLossPerSample = 0.053484067; EvalErrPerSample = 0.018139999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.273
-Starting Epoch 113: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[113 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05144035; EvalErr[0]PerSample = 0.01714844; TotalTime = 11.5189s; SamplesPerSecond = 2222.4
-Finished Epoch[113 of 160]: [Training Set] TrainLossPerSample = 0.053684823; EvalErrPerSample = 0.018139999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.298
-Starting Epoch 114: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[114 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05280342; EvalErr[0]PerSample = 0.01742188; TotalTime = 11.5954s; SamplesPerSecond = 2207.8
-Finished Epoch[114 of 160]: [Training Set] TrainLossPerSample = 0.05435824; EvalErrPerSample = 0.0184; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.4052
-Starting Epoch 115: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[115 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05222118; EvalErr[0]PerSample = 0.01757813; TotalTime = 11.6020s; SamplesPerSecond = 2206.5
-Finished Epoch[115 of 160]: [Training Set] TrainLossPerSample = 0.05275812; EvalErrPerSample = 0.0177; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3742
-Starting Epoch 116: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[116 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05418136; EvalErr[0]PerSample = 0.01832031; TotalTime = 11.5462s; SamplesPerSecond = 2217.2
-Finished Epoch[116 of 160]: [Training Set] TrainLossPerSample = 0.05442594; EvalErrPerSample = 0.018379999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3169
-Starting Epoch 117: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[117 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05089517; EvalErr[0]PerSample = 0.01753906; TotalTime = 11.4854s; SamplesPerSecond = 2228.9
-Finished Epoch[117 of 160]: [Training Set] TrainLossPerSample = 0.051556218; EvalErrPerSample = 0.018059999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2698
-Starting Epoch 118: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[118 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04923437; EvalErr[0]PerSample = 0.01722656; TotalTime = 11.6094s; SamplesPerSecond = 2205.1
-Finished Epoch[118 of 160]: [Training Set] TrainLossPerSample = 0.051423203; EvalErrPerSample = 0.01756; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3705
-Starting Epoch 119: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[119 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04878937; EvalErr[0]PerSample = 0.01687500; TotalTime = 11.5243s; SamplesPerSecond = 2221.4
-Finished Epoch[119 of 160]: [Training Set] TrainLossPerSample = 0.049813163; EvalErrPerSample = 0.01692; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3001
-Starting Epoch 120: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[120 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04832996; EvalErr[0]PerSample = 0.01691406; TotalTime = 11.4980s; SamplesPerSecond = 2226.5
-Finished Epoch[120 of 160]: [Training Set] TrainLossPerSample = 0.049110278; EvalErrPerSample = 0.01716; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3303
-Starting Epoch 121: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[121 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04482439; EvalErr[0]PerSample = 0.01488281; TotalTime = 11.5958s; SamplesPerSecond = 2207.7
-Finished Epoch[121 of 160]: [Training Set] TrainLossPerSample = 0.043392066; EvalErrPerSample = 0.01438; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4054
-Starting Epoch 122: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[122 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04098450; EvalErr[0]PerSample = 0.01304687; TotalTime = 11.5422s; SamplesPerSecond = 2217.9
-Finished Epoch[122 of 160]: [Training Set] TrainLossPerSample = 0.040084258; EvalErrPerSample = 0.0129; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4305
-Starting Epoch 123: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[123 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03640993; EvalErr[0]PerSample = 0.01175781; TotalTime = 11.5831s; SamplesPerSecond = 2210.1
-Finished Epoch[123 of 160]: [Training Set] TrainLossPerSample = 0.03892789; EvalErrPerSample = 0.01236; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3743
-Starting Epoch 124: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[124 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03714126; EvalErr[0]PerSample = 0.01171875; TotalTime = 11.5510s; SamplesPerSecond = 2216.3
-Finished Epoch[124 of 160]: [Training Set] TrainLossPerSample = 0.038824685; EvalErrPerSample = 0.01272; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3909
-Starting Epoch 125: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[125 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03836004; EvalErr[0]PerSample = 0.01242188; TotalTime = 11.5348s; SamplesPerSecond = 2219.4
-Finished Epoch[125 of 160]: [Training Set] TrainLossPerSample = 0.03861165; EvalErrPerSample = 0.01212; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2952
-Starting Epoch 126: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[126 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03456074; EvalErr[0]PerSample = 0.01113281; TotalTime = 11.5272s; SamplesPerSecond = 2220.8
-Finished Epoch[126 of 160]: [Training Set] TrainLossPerSample = 0.035167199; EvalErrPerSample = 0.01126; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2908
-Starting Epoch 127: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[127 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03603784; EvalErr[0]PerSample = 0.01101562; TotalTime = 11.3989s; SamplesPerSecond = 2245.8
-Finished Epoch[127 of 160]: [Training Set] TrainLossPerSample = 0.035970744; EvalErrPerSample = 0.01098; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.226
-Starting Epoch 128: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[128 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03553395; EvalErr[0]PerSample = 0.01175781; TotalTime = 11.5519s; SamplesPerSecond = 2216.1
-Finished Epoch[128 of 160]: [Training Set] TrainLossPerSample = 0.035689104; EvalErrPerSample = 0.01156; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3882
-Starting Epoch 129: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[129 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03523235; EvalErr[0]PerSample = 0.01109375; TotalTime = 11.5823s; SamplesPerSecond = 2210.3
-Finished Epoch[129 of 160]: [Training Set] TrainLossPerSample = 0.036777027; EvalErrPerSample = 0.01176; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3416
-Starting Epoch 130: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[130 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03563953; EvalErr[0]PerSample = 0.01058594; TotalTime = 11.4850s; SamplesPerSecond = 2229.0
-Finished Epoch[130 of 160]: [Training Set] TrainLossPerSample = 0.034818619; EvalErrPerSample = 0.010559999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2438
-Starting Epoch 131: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[131 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03574152; EvalErr[0]PerSample = 0.01187500; TotalTime = 11.4836s; SamplesPerSecond = 2229.3
-Finished Epoch[131 of 160]: [Training Set] TrainLossPerSample = 0.035892151; EvalErrPerSample = 0.01174; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2433
-Starting Epoch 132: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[132 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03485843; EvalErr[0]PerSample = 0.01152344; TotalTime = 11.4668s; SamplesPerSecond = 2232.5
-Finished Epoch[132 of 160]: [Training Set] TrainLossPerSample = 0.035623327; EvalErrPerSample = 0.01148; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2302
-Starting Epoch 133: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[133 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03582706; EvalErr[0]PerSample = 0.01058594; TotalTime = 11.4449s; SamplesPerSecond = 2236.8
-Finished Epoch[133 of 160]: [Training Set] TrainLossPerSample = 0.035650004; EvalErrPerSample = 0.01076; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2876
-Starting Epoch 134: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[134 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03308557; EvalErr[0]PerSample = 0.01031250; TotalTime = 11.5628s; SamplesPerSecond = 2214.0
-Finished Epoch[134 of 160]: [Training Set] TrainLossPerSample = 0.034471795; EvalErrPerSample = 0.010679999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3507
-Starting Epoch 135: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[135 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03313774; EvalErr[0]PerSample = 0.01011719; TotalTime = 11.5699s; SamplesPerSecond = 2212.6
-Finished Epoch[135 of 160]: [Training Set] TrainLossPerSample = 0.03418766; EvalErrPerSample = 0.01052; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.356
-Starting Epoch 136: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[136 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03477723; EvalErr[0]PerSample = 0.01121094; TotalTime = 11.5129s; SamplesPerSecond = 2223.6
-Finished Epoch[136 of 160]: [Training Set] TrainLossPerSample = 0.0342177; EvalErrPerSample = 0.01064; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2738
-Starting Epoch 137: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[137 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03409709; EvalErr[0]PerSample = 0.01007813; TotalTime = 11.4740s; SamplesPerSecond = 2231.1
-Finished Epoch[137 of 160]: [Training Set] TrainLossPerSample = 0.034409851; EvalErrPerSample = 0.01076; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2606
-Starting Epoch 138: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[138 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03408328; EvalErr[0]PerSample = 0.01019531; TotalTime = 11.6197s; SamplesPerSecond = 2203.1
-Finished Epoch[138 of 160]: [Training Set] TrainLossPerSample = 0.03398142; EvalErrPerSample = 0.01034; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4016
-Starting Epoch 139: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[139 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03391739; EvalErr[0]PerSample = 0.01003906; TotalTime = 11.5855s; SamplesPerSecond = 2209.7
-Finished Epoch[139 of 160]: [Training Set] TrainLossPerSample = 0.034369528; EvalErrPerSample = 0.0105; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4288
-Starting Epoch 140: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[140 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03432070; EvalErr[0]PerSample = 0.01093750; TotalTime = 11.5739s; SamplesPerSecond = 2211.9
-Finished Epoch[140 of 160]: [Training Set] TrainLossPerSample = 0.034433346; EvalErrPerSample = 0.01072; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3479
-Starting Epoch 141: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[141 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03326862; EvalErr[0]PerSample = 0.01023438; TotalTime = 11.5483s; SamplesPerSecond = 2216.8
-Finished Epoch[141 of 160]: [Training Set] TrainLossPerSample = 0.03355087; EvalErrPerSample = 0.01038; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.326
-Starting Epoch 142: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[142 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03197437; EvalErr[0]PerSample = 0.00988281; TotalTime = 11.4156s; SamplesPerSecond = 2242.6
-Finished Epoch[142 of 160]: [Training Set] TrainLossPerSample = 0.032701213; EvalErrPerSample = 0.0098000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.1816
-Starting Epoch 143: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[143 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03477237; EvalErr[0]PerSample = 0.01089844; TotalTime = 11.5553s; SamplesPerSecond = 2215.4
-Finished Epoch[143 of 160]: [Training Set] TrainLossPerSample = 0.034088843; EvalErrPerSample = 0.01066; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3259
-Starting Epoch 144: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[144 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03017725; EvalErr[0]PerSample = 0.00953125; TotalTime = 11.5897s; SamplesPerSecond = 2208.9
-Finished Epoch[144 of 160]: [Training Set] TrainLossPerSample = 0.031698398; EvalErrPerSample = 0.0099999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.5923
-Starting Epoch 145: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[145 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03448967; EvalErr[0]PerSample = 0.01074219; TotalTime = 11.4235s; SamplesPerSecond = 2241.0
-Finished Epoch[145 of 160]: [Training Set] TrainLossPerSample = 0.033083703; EvalErrPerSample = 0.01002; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.1806
-Starting Epoch 146: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[146 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03416681; EvalErr[0]PerSample = 0.01136719; TotalTime = 11.3855s; SamplesPerSecond = 2248.5
-Finished Epoch[146 of 160]: [Training Set] TrainLossPerSample = 0.034353275; EvalErrPerSample = 0.010799999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.1946
-Starting Epoch 147: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[147 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03433434; EvalErr[0]PerSample = 0.01101562; TotalTime = 11.5644s; SamplesPerSecond = 2213.7
-Finished Epoch[147 of 160]: [Training Set] TrainLossPerSample = 0.034649722; EvalErrPerSample = 0.011279999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3273
-Starting Epoch 148: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[148 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03219655; EvalErr[0]PerSample = 0.00988281; TotalTime = 11.5637s; SamplesPerSecond = 2213.8
-Finished Epoch[148 of 160]: [Training Set] TrainLossPerSample = 0.031285692; EvalErrPerSample = 0.0094999997; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3287
-Starting Epoch 149: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[149 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03190110; EvalErr[0]PerSample = 0.01046875; TotalTime = 11.4602s; SamplesPerSecond = 2233.8
-Finished Epoch[149 of 160]: [Training Set] TrainLossPerSample = 0.031930499; EvalErrPerSample = 0.0101; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2852
-Starting Epoch 150: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[150 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03089277; EvalErr[0]PerSample = 0.00921875; TotalTime = 11.5981s; SamplesPerSecond = 2207.3
-Finished Epoch[150 of 160]: [Training Set] TrainLossPerSample = 0.031741869; EvalErrPerSample = 0.00954; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4059
-Starting Epoch 151: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[151 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03341821; EvalErr[0]PerSample = 0.01105469; TotalTime = 11.5961s; SamplesPerSecond = 2207.6
-Finished Epoch[151 of 160]: [Training Set] TrainLossPerSample = 0.032794204; EvalErrPerSample = 0.01036; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.435
-Starting Epoch 152: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[152 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03127065; EvalErr[0]PerSample = 0.00937500; TotalTime = 11.5746s; SamplesPerSecond = 2211.7
-Finished Epoch[152 of 160]: [Training Set] TrainLossPerSample = 0.032228082; EvalErrPerSample = 0.0098599996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3883
-Starting Epoch 153: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[153 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03200430; EvalErr[0]PerSample = 0.01007813; TotalTime = 11.5357s; SamplesPerSecond = 2219.2
-Finished Epoch[153 of 160]: [Training Set] TrainLossPerSample = 0.031472486; EvalErrPerSample = 0.0098999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2992
-Starting Epoch 154: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[154 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03112781; EvalErr[0]PerSample = 0.00890625; TotalTime = 11.4339s; SamplesPerSecond = 2239.0
-Finished Epoch[154 of 160]: [Training Set] TrainLossPerSample = 0.030955072; EvalErrPerSample = 0.00954; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2657
-Starting Epoch 155: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[155 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03170617; EvalErr[0]PerSample = 0.00941406; TotalTime = 11.5475s; SamplesPerSecond = 2216.9
-Finished Epoch[155 of 160]: [Training Set] TrainLossPerSample = 0.031180983; EvalErrPerSample = 0.0093200002; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3148
-Starting Epoch 156: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[156 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02884187; EvalErr[0]PerSample = 0.00929687; TotalTime = 11.4932s; SamplesPerSecond = 2227.4
-Finished Epoch[156 of 160]: [Training Set] TrainLossPerSample = 0.030790305; EvalErrPerSample = 0.0097199995; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3262
-Starting Epoch 157: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[157 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03166680; EvalErr[0]PerSample = 0.00992188; TotalTime = 11.6570s; SamplesPerSecond = 2196.1
-Finished Epoch[157 of 160]: [Training Set] TrainLossPerSample = 0.031061091; EvalErrPerSample = 0.0094599994; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4556
-Starting Epoch 158: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[158 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03136421; EvalErr[0]PerSample = 0.00906250; TotalTime = 11.5571s; SamplesPerSecond = 2215.1
-Finished Epoch[158 of 160]: [Training Set] TrainLossPerSample = 0.029945096; EvalErrPerSample = 0.0084999995; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3185
-Starting Epoch 159: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[159 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03113249; EvalErr[0]PerSample = 0.00976563; TotalTime = 11.5817s; SamplesPerSecond = 2210.4
-Finished Epoch[159 of 160]: [Training Set] TrainLossPerSample = 0.03114122; EvalErrPerSample = 0.0093599996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.343
-Starting Epoch 160: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[160 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03009957; EvalErr[0]PerSample = 0.00886719; TotalTime = 11.5242s; SamplesPerSecond = 2221.4
-Finished Epoch[160 of 160]: [Training Set] TrainLossPerSample = 0.03085361; EvalErrPerSample = 0.00942; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3018
-CNTKCommandTrainEnd: Train
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
-
-Post-processing network...
-
-3 roots:
-	OutputNodes.z = Plus
-	Err = ErrorPrediction
-	CE = CrossEntropyWithSoftmax
-FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
-FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
-FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
-
-
-Validating for node OutputNodes.z. 185 nodes to process in pass 1.
-
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-
-Validating for node OutputNodes.z. 77 nodes to process in pass 2.
-
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-
-Validating for node OutputNodes.z, final verification.
-
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-
-107 out of 185 nodes do not share the minibatch layout with the input data.
-
-
-Validating for node Err. 187 nodes to process in pass 1.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node Err. 78 nodes to process in pass 2.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node Err, final verification.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-108 out of 187 nodes do not share the minibatch layout with the input data.
-
-
-Validating for node CE. 187 nodes to process in pass 1.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node CE. 78 nodes to process in pass 2.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node CE, final verification.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-108 out of 187 nodes do not share the minibatch layout with the input data.
-
-Post-processing network complete.
-
-Post-processing network...
-
-3 roots:
-	OutputNodes.z = Plus
-	Err = ErrorPrediction
-	CE = CrossEntropyWithSoftmax
-FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
-FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
-FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
-
-
-Validating for node OutputNodes.z. 185 nodes to process in pass 1.
-
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-
-Validating for node OutputNodes.z. 77 nodes to process in pass 2.
-
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-
-Validating for node OutputNodes.z, final verification.
-
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-
-107 out of 185 nodes do not share the minibatch layout with the input data.
-
-
-Validating for node Err. 187 nodes to process in pass 1.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node Err. 78 nodes to process in pass 2.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node Err, final verification.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-108 out of 187 nodes do not share the minibatch layout with the input data.
-
-
-Validating for node CE. 187 nodes to process in pass 1.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node CE. 78 nodes to process in pass 2.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-Validating for node CE, final verification.
-
-Validating --> labels = InputValue -> [10 [10], MBSize 0]
-Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
-Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
-Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
-Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
-Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
-Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
-Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
-Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
-Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
-Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
-Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
- -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
-Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
-Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
-Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
-
-108 out of 187 nodes do not share the minibatch layout with the input data.
-
-Post-processing network complete.
-evalNodeNames are not specified, using all the default evalnodes and training criterion nodes.
-
-
-Allocating matrices for forward and/or backward propagation.
-Minibatch[1-20]: Samples Seen = 10000    Err: ErrorPrediction/Sample = 0.0819    CE: CrossEntropyWithSoftmax/Sample = 0.35141698    
-Final Results: Minibatch[1-20]: Samples Seen = 10000    Err: ErrorPrediction/Sample = 0.0819    CE: CrossEntropyWithSoftmax/Sample = 0.35141698    Perplexity = 1.4210798    
-COMPLETED
+-------------------------------------------------------------------
+Build info: 
+
+		Built time: Jan 12 2016 14:46:20
+		Last modified date: Mon Jan 11 11:39:54 2016
+		CUDA_PATH: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.0
+		Build Branch: 
+		Build SHA1: 
+		Built by alexeyk on z840-01           
+		Build Path: C:\src\cntk\Source\CNTK\
+-------------------------------------------------------------------
+running on z840-01 at 2016/01/14 10:36:01
+command line: 
+..\..\..\..\x64\Release\CNTK.exe configFile=03_ResNet.config 
+
+>>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>
+RootDir = "."
+ConfigDir = "$RootDir$"
+DataDir = "$RootDir$"
+OutputDir = "$RootDir$/Output"
+ModelDir = "$OutputDir$/Models"
+ndlMacros=$ConfigDir$/Macros.ndl
+precision=float
+deviceId=Auto
+prefetch=true
+parallelTrain=false
+command=Train:AddBNEval:Test
+stderr=$OutputDir$/03_ResNet
+traceLevel=1
+numMBsToShowResult=200
+Proj16to32Filename = $ConfigDir$/16to32.txt
+Proj32to64Filename = $ConfigDir$/32to64.txt
+Train=[
+    action=train
+    modelPath=$ModelDir$/03_ResNet
+     NDLNetworkBuilder=[
+        networkDescription=$ConfigDir$/03_ResNet.ndl
+    ]
+    SGD=[
+        epochSize=0
+        minibatchSize=128
+        learningRatesPerMB=1.0*80:0.1*40:0.01
+        momentumPerMB=0.9
+        maxEpochs=160
+        L2RegWeight=0.0001
+        dropoutRate=0
+        ParallelTrain=[
+            parallelizationMethod=DataParallelSGD
+            distributedMBReading=true
+            parallelizationStartEpoch=1
+            DataParallelSGD=[
+                gradientBits=32
+            ]
+        ]
+    ]
+    reader=[
+        readerType=ImageReader
+        file=$DataDir$/train_map.txt
+        randomize=Auto
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType=Random
+            cropRatio=0.8
+            jitterType=UniRatio
+            interpolations=Linear
+            meanFile=$ConfigDir$/CIFAR-10_mean.xml
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+AddBNEval=[    
+    action=edit
+    CurModel=$ModelDir$/03_ResNet
+    NewModel=$ModelDir$/03_ResNet.Eval
+    editPath=$ConfigDir$/03_ResNet.mel
+]
+Test=[
+    action=test
+    modelPath=$ModelDir$/03_ResNet.Eval
+    minibatchSize=512
+     NDLNetworkBuilder=[
+        networkDescription=$ConfigDir$/03_ResNet.ndl
+    ]
+    reader=[
+        readerType=ImageReader
+        file=$DataDir$/test_map.txt
+        randomize=Auto
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType=Center
+            cropRatio=1
+            jitterType=UniRatio
+            interpolations=Linear
+            meanFile=$ConfigDir$/CIFAR-10_mean.xml
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+
+<<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<
+
+>>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
+RootDir = "."
+ConfigDir = "."
+DataDir = "."
+OutputDir = "./Output"
+ModelDir = "./Output/Models"
+ndlMacros=./Macros.ndl
+precision=float
+deviceId=Auto
+prefetch=true
+parallelTrain=false
+command=Train:AddBNEval:Test
+stderr=./Output/03_ResNet
+traceLevel=1
+numMBsToShowResult=200
+Proj16to32Filename = ./16to32.txt
+Proj32to64Filename = ./32to64.txt
+Train=[
+    action=train
+    modelPath=./Output/Models/03_ResNet
+     NDLNetworkBuilder=[
+        networkDescription=./03_ResNet.ndl
+    ]
+    SGD=[
+        epochSize=0
+        minibatchSize=128
+        learningRatesPerMB=1.0*80:0.1*40:0.01
+        momentumPerMB=0.9
+        maxEpochs=160
+        L2RegWeight=0.0001
+        dropoutRate=0
+        ParallelTrain=[
+            parallelizationMethod=DataParallelSGD
+            distributedMBReading=true
+            parallelizationStartEpoch=1
+            DataParallelSGD=[
+                gradientBits=32
+            ]
+        ]
+    ]
+    reader=[
+        readerType=ImageReader
+        file=./train_map.txt
+        randomize=Auto
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType=Random
+            cropRatio=0.8
+            jitterType=UniRatio
+            interpolations=Linear
+            meanFile=./CIFAR-10_mean.xml
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+AddBNEval=[    
+    action=edit
+    CurModel=./Output/Models/03_ResNet
+    NewModel=./Output/Models/03_ResNet.Eval
+    editPath=./03_ResNet.mel
+]
+Test=[
+    action=test
+    modelPath=./Output/Models/03_ResNet.Eval
+    minibatchSize=512
+     NDLNetworkBuilder=[
+        networkDescription=./03_ResNet.ndl
+    ]
+    reader=[
+        readerType=ImageReader
+        file=./test_map.txt
+        randomize=Auto
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType=Center
+            cropRatio=1
+            jitterType=UniRatio
+            interpolations=Linear
+            meanFile=./CIFAR-10_mean.xml
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+
+<<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
+
+>>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
+configparameters: 03_ResNet.config:AddBNEval=[    
+    action=edit
+    CurModel=./Output/Models/03_ResNet
+    NewModel=./Output/Models/03_ResNet.Eval
+    editPath=./03_ResNet.mel
+]
+
+configparameters: 03_ResNet.config:command=Train:AddBNEval:Test
+configparameters: 03_ResNet.config:ConfigDir=.
+configparameters: 03_ResNet.config:DataDir=.
+configparameters: 03_ResNet.config:deviceId=Auto
+configparameters: 03_ResNet.config:ModelDir=./Output/Models
+configparameters: 03_ResNet.config:ndlMacros=./Macros.ndl
+configparameters: 03_ResNet.config:numMBsToShowResult=200
+configparameters: 03_ResNet.config:OutputDir=./Output
+configparameters: 03_ResNet.config:parallelTrain=false
+configparameters: 03_ResNet.config:precision=float
+configparameters: 03_ResNet.config:prefetch=true
+configparameters: 03_ResNet.config:Proj16to32Filename=./16to32.txt
+configparameters: 03_ResNet.config:Proj32to64Filename=./32to64.txt
+configparameters: 03_ResNet.config:RootDir=.
+configparameters: 03_ResNet.config:stderr=./Output/03_ResNet
+configparameters: 03_ResNet.config:Test=[
+    action=test
+    modelPath=./Output/Models/03_ResNet.Eval
+    minibatchSize=512
+     NDLNetworkBuilder=[
+        networkDescription=./03_ResNet.ndl
+    ]
+    reader=[
+        readerType=ImageReader
+        file=./test_map.txt
+        randomize=Auto
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType=Center
+            cropRatio=1
+            jitterType=UniRatio
+            interpolations=Linear
+            meanFile=./CIFAR-10_mean.xml
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+
+configparameters: 03_ResNet.config:traceLevel=1
+configparameters: 03_ResNet.config:Train=[
+    action=train
+    modelPath=./Output/Models/03_ResNet
+     NDLNetworkBuilder=[
+        networkDescription=./03_ResNet.ndl
+    ]
+    SGD=[
+        epochSize=0
+        minibatchSize=128
+        learningRatesPerMB=1.0*80:0.1*40:0.01
+        momentumPerMB=0.9
+        maxEpochs=160
+        L2RegWeight=0.0001
+        dropoutRate=0
+        ParallelTrain=[
+            parallelizationMethod=DataParallelSGD
+            distributedMBReading=true
+            parallelizationStartEpoch=1
+            DataParallelSGD=[
+                gradientBits=32
+            ]
+        ]
+    ]
+    reader=[
+        readerType=ImageReader
+        file=./train_map.txt
+        randomize=Auto
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType=Random
+            cropRatio=0.8
+            jitterType=UniRatio
+            interpolations=Linear
+            meanFile=./CIFAR-10_mean.xml
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+
+<<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
+command: Train AddBNEval Test 
+precision = float
+CNTKModelPath: ./Output/Models/03_ResNet
+CNTKCommandTrainInfo: Train : 160
+CNTKCommandTrainInfo: CNTKNoMoreCommands_Total : 160
+CNTKCommandTrainBegin: Train
+LockDevice: Locked GPU 0 to test availability.
+LockDevice: Unlocked GPU 0 after testing.
+LockDevice: Locked GPU 2 to test availability.
+LockDevice: Unlocked GPU 2 after testing.
+LockDevice: Locked GPU 1 to test availability.
+LockDevice: Unlocked GPU 1 after testing.
+LockDevice: Locked GPU 0 for exclusive use.
+NDLBuilder Using GPU 0
+Microsoft::MSR::CNTK::GPUMatrix<ElemType>::SetGaussianRandomValue (GPU): creating curand object with seed 1, sizeof(ElemType)==4
+
+Post-processing network...
+
+3 roots:
+	Err = ErrorPrediction
+	OutputNodes.z = Plus
+	CE = CrossEntropyWithSoftmax
+FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
+FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
+FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
+
+
+Validating for node Err. 187 nodes to process in pass 1.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node Err. 78 nodes to process in pass 2.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node Err, final verification.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+108 out of 187 nodes do not share the minibatch layout with the input data.
+
+
+Validating for node OutputNodes.z. 185 nodes to process in pass 1.
+
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+
+Validating for node OutputNodes.z. 77 nodes to process in pass 2.
+
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+
+Validating for node OutputNodes.z, final verification.
+
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+
+107 out of 185 nodes do not share the minibatch layout with the input data.
+
+
+Validating for node CE. 187 nodes to process in pass 1.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node CE. 78 nodes to process in pass 2.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node CE, final verification.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+108 out of 187 nodes do not share the minibatch layout with the input data.
+
+Post-processing network complete.
+
+SGD using GPU 0.
+
+Training criterion node(s):
+	CE = CrossEntropyWithSoftmax
+
+Evaluation criterion node(s):
+	Err = ErrorPrediction
+
+
+Allocating matrices for forward and/or backward propagation.
+No PreCompute nodes found, skipping PreCompute step
+Set Max Temp Mem Size For Convolution Nodes to 0 samples.
+Starting Epoch 1: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+#NLop10#
+Tensor Op: Op 10: 32 x 32 x 16 x 128 x 1 -> 32 x 32 x 16 x 128 x 1
+24 procs  32 warps  2147483647 65535 65535 max grid  on  GeForce GTX TITAN X
+3 procs  32 warps  2147483647 65535 65535 max grid  on  Quadro K620
+14 procs  32 warps  2147483647 65535 65535 max grid  on  GeForce GTX TITAN
+Tensor Op: Op 15: 32 x 32 x 16 x 128 x 1 op 32 x 32 x 16 x 128 x 1 -> 32 x 32 x 16 x 128 x 1
+ Epoch[ 1 of 160]-Minibatch[   1- 200]: SamplesSeen = 25600; TrainLossPerSample =  1.78511215; EvalErr[0]PerSample = 0.67218750; TotalTime = 15.4090s; SamplesPerSecond = 1661.4
+Finished Epoch[ 1 of 160]: [Training Set] TrainLossPerSample = 1.6011882; EvalErrPerSample = 0.59435999; AvgLearningRatePerSample = 0.0078125; EpochTime=26.129
+Starting Epoch 2: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 2 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  1.18311264; EvalErr[0]PerSample = 0.42843750; TotalTime = 11.4034s; SamplesPerSecond = 2245.0
+Finished Epoch[ 2 of 160]: [Training Set] TrainLossPerSample = 1.1033459; EvalErrPerSample = 0.3969; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1287
+Starting Epoch 3: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 3 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.90734154; EvalErr[0]PerSample = 0.32207031; TotalTime = 11.3750s; SamplesPerSecond = 2250.5
+Finished Epoch[ 3 of 160]: [Training Set] TrainLossPerSample = 0.86615896; EvalErrPerSample = 0.30719998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.0657
+Starting Epoch 4: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 4 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.77141624; EvalErr[0]PerSample = 0.27066406; TotalTime = 11.4567s; SamplesPerSecond = 2234.5
+Finished Epoch[ 4 of 160]: [Training Set] TrainLossPerSample = 0.75146842; EvalErrPerSample = 0.26264; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2146
+Starting Epoch 5: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 5 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.68756187; EvalErr[0]PerSample = 0.23796875; TotalTime = 11.4042s; SamplesPerSecond = 2244.8
+Finished Epoch[ 5 of 160]: [Training Set] TrainLossPerSample = 0.67170274; EvalErrPerSample = 0.23289999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1571
+Starting Epoch 6: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 6 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.61656025; EvalErr[0]PerSample = 0.21398438; TotalTime = 11.4800s; SamplesPerSecond = 2230.0
+Finished Epoch[ 6 of 160]: [Training Set] TrainLossPerSample = 0.612014; EvalErrPerSample = 0.21263999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2553
+Starting Epoch 7: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 7 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.58124241; EvalErr[0]PerSample = 0.20335938; TotalTime = 11.4640s; SamplesPerSecond = 2233.1
+Finished Epoch[ 7 of 160]: [Training Set] TrainLossPerSample = 0.56962705; EvalErrPerSample = 0.19909999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2453
+Starting Epoch 8: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 8 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.54073845; EvalErr[0]PerSample = 0.18796875; TotalTime = 11.5047s; SamplesPerSecond = 2225.2
+Finished Epoch[ 8 of 160]: [Training Set] TrainLossPerSample = 0.53796959; EvalErrPerSample = 0.18616; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3345
+Starting Epoch 9: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 9 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.50834869; EvalErr[0]PerSample = 0.17632813; TotalTime = 11.5167s; SamplesPerSecond = 2222.9
+Finished Epoch[ 9 of 160]: [Training Set] TrainLossPerSample = 0.51177925; EvalErrPerSample = 0.17704; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3221
+Starting Epoch 10: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[10 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.48615803; EvalErr[0]PerSample = 0.16917969; TotalTime = 11.4815s; SamplesPerSecond = 2229.7
+Finished Epoch[10 of 160]: [Training Set] TrainLossPerSample = 0.49343586; EvalErrPerSample = 0.17158; AvgLearningRatePerSample = 0.0078125; EpochTime=22.286
+Starting Epoch 11: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[11 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.46868595; EvalErr[0]PerSample = 0.16312500; TotalTime = 11.4115s; SamplesPerSecond = 2243.4
+Finished Epoch[11 of 160]: [Training Set] TrainLossPerSample = 0.46725979; EvalErrPerSample = 0.16214; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2195
+Starting Epoch 12: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[12 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.45617180; EvalErr[0]PerSample = 0.15531250; TotalTime = 11.4316s; SamplesPerSecond = 2239.4
+Finished Epoch[12 of 160]: [Training Set] TrainLossPerSample = 0.44944596; EvalErrPerSample = 0.15497999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2204
+Starting Epoch 13: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[13 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.43442238; EvalErr[0]PerSample = 0.15031250; TotalTime = 11.5970s; SamplesPerSecond = 2207.5
+Finished Epoch[13 of 160]: [Training Set] TrainLossPerSample = 0.43854889; EvalErrPerSample = 0.15235999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4304
+Starting Epoch 14: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[14 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.41474213; EvalErr[0]PerSample = 0.14265625; TotalTime = 11.5551s; SamplesPerSecond = 2215.5
+Finished Epoch[14 of 160]: [Training Set] TrainLossPerSample = 0.42394838; EvalErrPerSample = 0.14659999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3929
+Starting Epoch 15: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[15 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.40827648; EvalErr[0]PerSample = 0.14167969; TotalTime = 11.5593s; SamplesPerSecond = 2214.7
+Finished Epoch[15 of 160]: [Training Set] TrainLossPerSample = 0.40738714; EvalErrPerSample = 0.14126; AvgLearningRatePerSample = 0.0078125; EpochTime=22.404
+Starting Epoch 16: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[16 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.39572418; EvalErr[0]PerSample = 0.13765625; TotalTime = 11.6490s; SamplesPerSecond = 2197.6
+Finished Epoch[16 of 160]: [Training Set] TrainLossPerSample = 0.4007735; EvalErrPerSample = 0.13798; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4156
+Starting Epoch 17: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[17 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.38931370; EvalErr[0]PerSample = 0.13757813; TotalTime = 11.4136s; SamplesPerSecond = 2242.9
+Finished Epoch[17 of 160]: [Training Set] TrainLossPerSample = 0.39249194; EvalErrPerSample = 0.13716; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1748
+Starting Epoch 18: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[18 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.37350758; EvalErr[0]PerSample = 0.12980469; TotalTime = 11.5431s; SamplesPerSecond = 2217.8
+Finished Epoch[18 of 160]: [Training Set] TrainLossPerSample = 0.37985951; EvalErrPerSample = 0.13271999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3177
+Starting Epoch 19: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[19 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.37046688; EvalErr[0]PerSample = 0.12742187; TotalTime = 11.4904s; SamplesPerSecond = 2228.0
+Finished Epoch[19 of 160]: [Training Set] TrainLossPerSample = 0.37240577; EvalErrPerSample = 0.12842; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2594
+Starting Epoch 20: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[20 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.36126266; EvalErr[0]PerSample = 0.12656250; TotalTime = 11.5121s; SamplesPerSecond = 2223.7
+Finished Epoch[20 of 160]: [Training Set] TrainLossPerSample = 0.36667159; EvalErrPerSample = 0.12833999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2756
+Starting Epoch 21: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[21 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.36016537; EvalErr[0]PerSample = 0.12242188; TotalTime = 11.5126s; SamplesPerSecond = 2223.7
+Finished Epoch[21 of 160]: [Training Set] TrainLossPerSample = 0.35836998; EvalErrPerSample = 0.12233999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2811
+Starting Epoch 22: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[22 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.34188179; EvalErr[0]PerSample = 0.11910156; TotalTime = 11.5167s; SamplesPerSecond = 2222.9
+Finished Epoch[22 of 160]: [Training Set] TrainLossPerSample = 0.34598714; EvalErrPerSample = 0.12013999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3214
+Starting Epoch 23: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[23 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.34518520; EvalErr[0]PerSample = 0.11894531; TotalTime = 11.5177s; SamplesPerSecond = 2222.7
+Finished Epoch[23 of 160]: [Training Set] TrainLossPerSample = 0.34407225; EvalErrPerSample = 0.11892; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3061
+Starting Epoch 24: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[24 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.33347424; EvalErr[0]PerSample = 0.11660156; TotalTime = 11.5769s; SamplesPerSecond = 2211.3
+Finished Epoch[24 of 160]: [Training Set] TrainLossPerSample = 0.34192094; EvalErrPerSample = 0.11848; AvgLearningRatePerSample = 0.0078125; EpochTime=22.368
+Starting Epoch 25: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[25 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.33844280; EvalErr[0]PerSample = 0.11648438; TotalTime = 11.5529s; SamplesPerSecond = 2215.9
+Finished Epoch[25 of 160]: [Training Set] TrainLossPerSample = 0.3405562; EvalErrPerSample = 0.11768; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3158
+Starting Epoch 26: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[26 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.32573421; EvalErr[0]PerSample = 0.11328125; TotalTime = 11.4885s; SamplesPerSecond = 2228.3
+Finished Epoch[26 of 160]: [Training Set] TrainLossPerSample = 0.3283667; EvalErrPerSample = 0.11268; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3241
+Starting Epoch 27: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[27 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.32139095; EvalErr[0]PerSample = 0.11300781; TotalTime = 11.5773s; SamplesPerSecond = 2211.2
+Finished Epoch[27 of 160]: [Training Set] TrainLossPerSample = 0.32817245; EvalErrPerSample = 0.11406; AvgLearningRatePerSample = 0.0078125; EpochTime=22.345
+Starting Epoch 28: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[28 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.31476902; EvalErr[0]PerSample = 0.11074219; TotalTime = 11.4481s; SamplesPerSecond = 2236.2
+Finished Epoch[28 of 160]: [Training Set] TrainLossPerSample = 0.32170638; EvalErrPerSample = 0.1133; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2168
+Starting Epoch 29: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[29 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30912992; EvalErr[0]PerSample = 0.10859375; TotalTime = 11.4067s; SamplesPerSecond = 2244.3
+Finished Epoch[29 of 160]: [Training Set] TrainLossPerSample = 0.31786552; EvalErrPerSample = 0.11086; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1891
+Starting Epoch 30: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[30 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30547693; EvalErr[0]PerSample = 0.10496094; TotalTime = 11.5656s; SamplesPerSecond = 2213.5
+Finished Epoch[30 of 160]: [Training Set] TrainLossPerSample = 0.31253323; EvalErrPerSample = 0.1078; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3465
+Starting Epoch 31: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[31 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30724499; EvalErr[0]PerSample = 0.10769531; TotalTime = 11.5380s; SamplesPerSecond = 2218.8
+Finished Epoch[31 of 160]: [Training Set] TrainLossPerSample = 0.31017771; EvalErrPerSample = 0.10802; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3988
+Starting Epoch 32: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[32 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30443895; EvalErr[0]PerSample = 0.10660156; TotalTime = 11.5474s; SamplesPerSecond = 2216.9
+Finished Epoch[32 of 160]: [Training Set] TrainLossPerSample = 0.31054172; EvalErrPerSample = 0.10764; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3142
+Starting Epoch 33: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[33 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.29248705; EvalErr[0]PerSample = 0.10175781; TotalTime = 11.4051s; SamplesPerSecond = 2244.6
+Finished Epoch[33 of 160]: [Training Set] TrainLossPerSample = 0.30296284; EvalErrPerSample = 0.10422; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1968
+Starting Epoch 34: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[34 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30263618; EvalErr[0]PerSample = 0.10550781; TotalTime = 11.5888s; SamplesPerSecond = 2209.0
+Finished Epoch[34 of 160]: [Training Set] TrainLossPerSample = 0.30190024; EvalErrPerSample = 0.10473999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4491
+Starting Epoch 35: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[35 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.29659115; EvalErr[0]PerSample = 0.10171875; TotalTime = 11.6140s; SamplesPerSecond = 2204.2
+Finished Epoch[35 of 160]: [Training Set] TrainLossPerSample = 0.29864648; EvalErrPerSample = 0.10306; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4411
+Starting Epoch 36: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[36 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28713980; EvalErr[0]PerSample = 0.10027344; TotalTime = 11.5617s; SamplesPerSecond = 2214.2
+Finished Epoch[36 of 160]: [Training Set] TrainLossPerSample = 0.29245624; EvalErrPerSample = 0.10234; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3455
+Starting Epoch 37: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[37 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28737648; EvalErr[0]PerSample = 0.09847656; TotalTime = 11.4851s; SamplesPerSecond = 2229.0
+Finished Epoch[37 of 160]: [Training Set] TrainLossPerSample = 0.29400381; EvalErrPerSample = 0.10157999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3035
+Starting Epoch 38: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[38 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28604475; EvalErr[0]PerSample = 0.09699219; TotalTime = 11.6237s; SamplesPerSecond = 2202.4
+Finished Epoch[38 of 160]: [Training Set] TrainLossPerSample = 0.29055047; EvalErrPerSample = 0.1006; AvgLearningRatePerSample = 0.0078125; EpochTime=22.41
+Starting Epoch 39: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[39 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28677174; EvalErr[0]PerSample = 0.09929688; TotalTime = 11.6037s; SamplesPerSecond = 2206.2
+Finished Epoch[39 of 160]: [Training Set] TrainLossPerSample = 0.28853956; EvalErrPerSample = 0.101; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4156
+Starting Epoch 40: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[40 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27731121; EvalErr[0]PerSample = 0.09699219; TotalTime = 11.5688s; SamplesPerSecond = 2212.9
+Finished Epoch[40 of 160]: [Training Set] TrainLossPerSample = 0.28800672; EvalErrPerSample = 0.099959999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3445
+Starting Epoch 41: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[41 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28877285; EvalErr[0]PerSample = 0.10156250; TotalTime = 11.5167s; SamplesPerSecond = 2222.9
+Finished Epoch[41 of 160]: [Training Set] TrainLossPerSample = 0.28422225; EvalErrPerSample = 0.098719999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.355
+Starting Epoch 42: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[42 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27758221; EvalErr[0]PerSample = 0.09710937; TotalTime = 11.5935s; SamplesPerSecond = 2208.1
+Finished Epoch[42 of 160]: [Training Set] TrainLossPerSample = 0.28059965; EvalErrPerSample = 0.097819999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.7852
+Starting Epoch 43: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[43 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27680891; EvalErr[0]PerSample = 0.09644531; TotalTime = 11.6198s; SamplesPerSecond = 2203.1
+Finished Epoch[43 of 160]: [Training Set] TrainLossPerSample = 0.28217828; EvalErrPerSample = 0.098200001; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4529
+Starting Epoch 44: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[44 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27107407; EvalErr[0]PerSample = 0.09421875; TotalTime = 11.5929s; SamplesPerSecond = 2208.2
+Finished Epoch[44 of 160]: [Training Set] TrainLossPerSample = 0.27688959; EvalErrPerSample = 0.096079998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3608
+Starting Epoch 45: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[45 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27329311; EvalErr[0]PerSample = 0.09585937; TotalTime = 11.4535s; SamplesPerSecond = 2235.1
+Finished Epoch[45 of 160]: [Training Set] TrainLossPerSample = 0.27860796; EvalErrPerSample = 0.097539999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2187
+Starting Epoch 46: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[46 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.26291889; EvalErr[0]PerSample = 0.09253906; TotalTime = 11.4315s; SamplesPerSecond = 2239.4
+Finished Epoch[46 of 160]: [Training Set] TrainLossPerSample = 0.27177998; EvalErrPerSample = 0.095179997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1909
+Starting Epoch 47: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[47 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27016382; EvalErr[0]PerSample = 0.09503906; TotalTime = 11.4016s; SamplesPerSecond = 2245.3
+Finished Epoch[47 of 160]: [Training Set] TrainLossPerSample = 0.27474955; EvalErrPerSample = 0.096859999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.1576
+Starting Epoch 48: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[48 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25722231; EvalErr[0]PerSample = 0.08960938; TotalTime = 11.4404s; SamplesPerSecond = 2237.7
+Finished Epoch[48 of 160]: [Training Set] TrainLossPerSample = 0.26707178; EvalErrPerSample = 0.09296; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2165
+Starting Epoch 49: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[49 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25658318; EvalErr[0]PerSample = 0.08960938; TotalTime = 11.5004s; SamplesPerSecond = 2226.0
+Finished Epoch[49 of 160]: [Training Set] TrainLossPerSample = 0.26653028; EvalErrPerSample = 0.093059994; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2617
+Starting Epoch 50: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[50 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.26309399; EvalErr[0]PerSample = 0.09062500; TotalTime = 11.4840s; SamplesPerSecond = 2229.2
+Finished Epoch[50 of 160]: [Training Set] TrainLossPerSample = 0.27242121; EvalErrPerSample = 0.094359994; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4308
+Starting Epoch 51: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[51 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25116308; EvalErr[0]PerSample = 0.08757813; TotalTime = 11.5958s; SamplesPerSecond = 2207.7
+Finished Epoch[51 of 160]: [Training Set] TrainLossPerSample = 0.26714918; EvalErrPerSample = 0.093719997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3675
+Starting Epoch 52: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[52 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25991890; EvalErr[0]PerSample = 0.09207031; TotalTime = 11.4533s; SamplesPerSecond = 2235.2
+Finished Epoch[52 of 160]: [Training Set] TrainLossPerSample = 0.2633768; EvalErrPerSample = 0.09268; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3111
+Starting Epoch 53: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[53 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25488058; EvalErr[0]PerSample = 0.08957031; TotalTime = 11.5890s; SamplesPerSecond = 2209.0
+Finished Epoch[53 of 160]: [Training Set] TrainLossPerSample = 0.26292121; EvalErrPerSample = 0.092160001; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3534
+Starting Epoch 54: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[54 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24893934; EvalErr[0]PerSample = 0.08859375; TotalTime = 11.5135s; SamplesPerSecond = 2223.5
+Finished Epoch[54 of 160]: [Training Set] TrainLossPerSample = 0.25717214; EvalErrPerSample = 0.090659998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4264
+Starting Epoch 55: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[55 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25861763; EvalErr[0]PerSample = 0.09019531; TotalTime = 11.6295s; SamplesPerSecond = 2201.3
+Finished Epoch[55 of 160]: [Training Set] TrainLossPerSample = 0.26373541; EvalErrPerSample = 0.091260001; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3938
+Starting Epoch 56: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[56 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25139614; EvalErr[0]PerSample = 0.08546875; TotalTime = 11.5303s; SamplesPerSecond = 2220.2
+Finished Epoch[56 of 160]: [Training Set] TrainLossPerSample = 0.25967881; EvalErrPerSample = 0.089499995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2997
+Starting Epoch 57: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[57 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24860460; EvalErr[0]PerSample = 0.08734375; TotalTime = 11.5494s; SamplesPerSecond = 2216.6
+Finished Epoch[57 of 160]: [Training Set] TrainLossPerSample = 0.25601548; EvalErrPerSample = 0.090059996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3412
+Starting Epoch 58: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[58 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24636944; EvalErr[0]PerSample = 0.08496094; TotalTime = 11.5079s; SamplesPerSecond = 2224.6
+Finished Epoch[58 of 160]: [Training Set] TrainLossPerSample = 0.25675642; EvalErrPerSample = 0.088599995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3309
+Starting Epoch 59: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[59 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25354239; EvalErr[0]PerSample = 0.09003906; TotalTime = 11.5633s; SamplesPerSecond = 2213.9
+Finished Epoch[59 of 160]: [Training Set] TrainLossPerSample = 0.25516802; EvalErrPerSample = 0.090879999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3313
+Starting Epoch 60: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[60 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24986008; EvalErr[0]PerSample = 0.08527344; TotalTime = 11.4490s; SamplesPerSecond = 2236.0
+Finished Epoch[60 of 160]: [Training Set] TrainLossPerSample = 0.25723642; EvalErrPerSample = 0.088979997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.242
+Starting Epoch 61: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[61 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24366058; EvalErr[0]PerSample = 0.08671875; TotalTime = 11.5741s; SamplesPerSecond = 2211.8
+Finished Epoch[61 of 160]: [Training Set] TrainLossPerSample = 0.25719753; EvalErrPerSample = 0.091239996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3397
+Starting Epoch 62: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[62 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24259378; EvalErr[0]PerSample = 0.08492187; TotalTime = 11.5269s; SamplesPerSecond = 2220.9
+Finished Epoch[62 of 160]: [Training Set] TrainLossPerSample = 0.2515817; EvalErrPerSample = 0.088679999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3014
+Starting Epoch 63: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[63 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24029449; EvalErr[0]PerSample = 0.08460938; TotalTime = 11.5659s; SamplesPerSecond = 2213.4
+Finished Epoch[63 of 160]: [Training Set] TrainLossPerSample = 0.25265002; EvalErrPerSample = 0.088959999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3815
+Starting Epoch 64: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[64 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23263992; EvalErr[0]PerSample = 0.08144531; TotalTime = 11.5963s; SamplesPerSecond = 2207.6
+Finished Epoch[64 of 160]: [Training Set] TrainLossPerSample = 0.24182168; EvalErrPerSample = 0.084859997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.4068
+Starting Epoch 65: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[65 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24372847; EvalErr[0]PerSample = 0.08410156; TotalTime = 11.5519s; SamplesPerSecond = 2216.1
+Finished Epoch[65 of 160]: [Training Set] TrainLossPerSample = 0.25128376; EvalErrPerSample = 0.087699994; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3194
+Starting Epoch 66: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[66 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23706863; EvalErr[0]PerSample = 0.08332031; TotalTime = 11.5141s; SamplesPerSecond = 2223.4
+Finished Epoch[66 of 160]: [Training Set] TrainLossPerSample = 0.2469063; EvalErrPerSample = 0.085879996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2777
+Starting Epoch 67: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[67 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23676601; EvalErr[0]PerSample = 0.08445313; TotalTime = 11.4886s; SamplesPerSecond = 2228.3
+Finished Epoch[67 of 160]: [Training Set] TrainLossPerSample = 0.2486082; EvalErrPerSample = 0.087859996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2538
+Starting Epoch 68: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[68 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23655962; EvalErr[0]PerSample = 0.08148437; TotalTime = 11.5050s; SamplesPerSecond = 2225.1
+Finished Epoch[68 of 160]: [Training Set] TrainLossPerSample = 0.24334124; EvalErrPerSample = 0.084699996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2913
+Starting Epoch 69: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[69 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23345692; EvalErr[0]PerSample = 0.08085937; TotalTime = 11.5400s; SamplesPerSecond = 2218.4
+Finished Epoch[69 of 160]: [Training Set] TrainLossPerSample = 0.241666; EvalErrPerSample = 0.083919995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3544
+Starting Epoch 70: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[70 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23493608; EvalErr[0]PerSample = 0.08160156; TotalTime = 11.5393s; SamplesPerSecond = 2218.5
+Finished Epoch[70 of 160]: [Training Set] TrainLossPerSample = 0.24537624; EvalErrPerSample = 0.084799998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3067
+Starting Epoch 71: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[71 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23666447; EvalErr[0]PerSample = 0.08300781; TotalTime = 11.5368s; SamplesPerSecond = 2219.0
+Finished Epoch[71 of 160]: [Training Set] TrainLossPerSample = 0.24771863; EvalErrPerSample = 0.086999997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3618
+Starting Epoch 72: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[72 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24186720; EvalErr[0]PerSample = 0.08511719; TotalTime = 11.5608s; SamplesPerSecond = 2214.4
+Finished Epoch[72 of 160]: [Training Set] TrainLossPerSample = 0.2442316; EvalErrPerSample = 0.085159995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3384
+Starting Epoch 73: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[73 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23745417; EvalErr[0]PerSample = 0.08406250; TotalTime = 11.5428s; SamplesPerSecond = 2217.8
+Finished Epoch[73 of 160]: [Training Set] TrainLossPerSample = 0.24105896; EvalErrPerSample = 0.084639996; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3139
+Starting Epoch 74: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[74 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25070684; EvalErr[0]PerSample = 0.08886719; TotalTime = 11.4398s; SamplesPerSecond = 2237.8
+Finished Epoch[74 of 160]: [Training Set] TrainLossPerSample = 0.24804311; EvalErrPerSample = 0.087299995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.2443
+Starting Epoch 75: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[75 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23875351; EvalErr[0]PerSample = 0.08117188; TotalTime = 11.5714s; SamplesPerSecond = 2212.3
+Finished Epoch[75 of 160]: [Training Set] TrainLossPerSample = 0.24082842; EvalErrPerSample = 0.083080001; AvgLearningRatePerSample = 0.0078125; EpochTime=22.366
+Starting Epoch 76: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[76 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23140110; EvalErr[0]PerSample = 0.08203125; TotalTime = 11.5559s; SamplesPerSecond = 2215.3
+Finished Epoch[76 of 160]: [Training Set] TrainLossPerSample = 0.24258973; EvalErrPerSample = 0.084919997; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3202
+Starting Epoch 77: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[77 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23099312; EvalErr[0]PerSample = 0.08105469; TotalTime = 11.5710s; SamplesPerSecond = 2212.4
+Finished Epoch[77 of 160]: [Training Set] TrainLossPerSample = 0.24171358; EvalErrPerSample = 0.084399998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.416
+Starting Epoch 78: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[78 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.22422340; EvalErr[0]PerSample = 0.07945312; TotalTime = 11.5828s; SamplesPerSecond = 2210.2
+Finished Epoch[78 of 160]: [Training Set] TrainLossPerSample = 0.23471047; EvalErrPerSample = 0.083219998; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3526
+Starting Epoch 79: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[79 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23049221; EvalErr[0]PerSample = 0.08292969; TotalTime = 11.5700s; SamplesPerSecond = 2212.6
+Finished Epoch[79 of 160]: [Training Set] TrainLossPerSample = 0.237709; EvalErrPerSample = 0.083939999; AvgLearningRatePerSample = 0.0078125; EpochTime=22.3438
+Starting Epoch 80: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[80 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23512028; EvalErr[0]PerSample = 0.08218750; TotalTime = 11.4967s; SamplesPerSecond = 2226.7
+Finished Epoch[80 of 160]: [Training Set] TrainLossPerSample = 0.23986502; EvalErrPerSample = 0.084139995; AvgLearningRatePerSample = 0.0078125; EpochTime=22.32
+Starting Epoch 81: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[81 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18993895; EvalErr[0]PerSample = 0.06554688; TotalTime = 11.5483s; SamplesPerSecond = 2216.8
+Finished Epoch[81 of 160]: [Training Set] TrainLossPerSample = 0.16793214; EvalErrPerSample = 0.057559997; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3134
+Starting Epoch 82: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[82 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.13101171; EvalErr[0]PerSample = 0.04542969; TotalTime = 11.5430s; SamplesPerSecond = 2217.8
+Finished Epoch[82 of 160]: [Training Set] TrainLossPerSample = 0.12739825; EvalErrPerSample = 0.04394; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3113
+Starting Epoch 83: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[83 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.11582967; EvalErr[0]PerSample = 0.03949219; TotalTime = 11.5821s; SamplesPerSecond = 2210.3
+Finished Epoch[83 of 160]: [Training Set] TrainLossPerSample = 0.11576424; EvalErrPerSample = 0.03926; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.4228
+Starting Epoch 84: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[84 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.10656901; EvalErr[0]PerSample = 0.03722656; TotalTime = 11.5453s; SamplesPerSecond = 2217.4
+Finished Epoch[84 of 160]: [Training Set] TrainLossPerSample = 0.106635; EvalErrPerSample = 0.03644; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.5034
+Starting Epoch 85: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[85 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.09979140; EvalErr[0]PerSample = 0.03359375; TotalTime = 11.4506s; SamplesPerSecond = 2235.7
+Finished Epoch[85 of 160]: [Training Set] TrainLossPerSample = 0.10094135; EvalErrPerSample = 0.033980001; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2631
+Starting Epoch 86: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[86 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.09689341; EvalErr[0]PerSample = 0.03265625; TotalTime = 11.5657s; SamplesPerSecond = 2213.4
+Finished Epoch[86 of 160]: [Training Set] TrainLossPerSample = 0.096568428; EvalErrPerSample = 0.033059999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3286
+Starting Epoch 87: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[87 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.09459194; EvalErr[0]PerSample = 0.03140625; TotalTime = 11.4668s; SamplesPerSecond = 2232.5
+Finished Epoch[87 of 160]: [Training Set] TrainLossPerSample = 0.09401314; EvalErrPerSample = 0.031119999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2404
+Starting Epoch 88: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[88 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.09019656; EvalErr[0]PerSample = 0.02929688; TotalTime = 11.4755s; SamplesPerSecond = 2230.8
+Finished Epoch[88 of 160]: [Training Set] TrainLossPerSample = 0.090820439; EvalErrPerSample = 0.030199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2533
+Starting Epoch 89: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[89 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.08303898; EvalErr[0]PerSample = 0.02785156; TotalTime = 11.5285s; SamplesPerSecond = 2220.6
+Finished Epoch[89 of 160]: [Training Set] TrainLossPerSample = 0.086739741; EvalErrPerSample = 0.02956; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.307
+Starting Epoch 90: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[90 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.08771500; EvalErr[0]PerSample = 0.02984375; TotalTime = 11.5321s; SamplesPerSecond = 2219.9
+Finished Epoch[90 of 160]: [Training Set] TrainLossPerSample = 0.087390222; EvalErrPerSample = 0.03018; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3574
+Starting Epoch 91: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[91 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.08021155; EvalErr[0]PerSample = 0.02699219; TotalTime = 11.5502s; SamplesPerSecond = 2216.4
+Finished Epoch[91 of 160]: [Training Set] TrainLossPerSample = 0.080674298; EvalErrPerSample = 0.027319999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3191
+Starting Epoch 92: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[92 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.08322118; EvalErr[0]PerSample = 0.02910156; TotalTime = 11.5210s; SamplesPerSecond = 2222.0
+Finished Epoch[92 of 160]: [Training Set] TrainLossPerSample = 0.08106219; EvalErrPerSample = 0.028339999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2938
+Starting Epoch 93: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[93 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.08073726; EvalErr[0]PerSample = 0.02789062; TotalTime = 11.3990s; SamplesPerSecond = 2245.8
+Finished Epoch[93 of 160]: [Training Set] TrainLossPerSample = 0.079686686; EvalErrPerSample = 0.027419999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2285
+Starting Epoch 94: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[94 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07666363; EvalErr[0]PerSample = 0.02695313; TotalTime = 11.5560s; SamplesPerSecond = 2215.3
+Finished Epoch[94 of 160]: [Training Set] TrainLossPerSample = 0.077968039; EvalErrPerSample = 0.027099999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3268
+Starting Epoch 95: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[95 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07550663; EvalErr[0]PerSample = 0.02609375; TotalTime = 11.4795s; SamplesPerSecond = 2230.1
+Finished Epoch[95 of 160]: [Training Set] TrainLossPerSample = 0.076805077; EvalErrPerSample = 0.02616; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2683
+Starting Epoch 96: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[96 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07483893; EvalErr[0]PerSample = 0.02652344; TotalTime = 11.4940s; SamplesPerSecond = 2227.3
+Finished Epoch[96 of 160]: [Training Set] TrainLossPerSample = 0.073900625; EvalErrPerSample = 0.025839999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2583
+Starting Epoch 97: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[97 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07033884; EvalErr[0]PerSample = 0.02421875; TotalTime = 11.4972s; SamplesPerSecond = 2226.6
+Finished Epoch[97 of 160]: [Training Set] TrainLossPerSample = 0.072241709; EvalErrPerSample = 0.025239998; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3373
+Starting Epoch 98: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[98 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07043020; EvalErr[0]PerSample = 0.02421875; TotalTime = 11.5606s; SamplesPerSecond = 2214.4
+Finished Epoch[98 of 160]: [Training Set] TrainLossPerSample = 0.070759542; EvalErrPerSample = 0.024279999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.328
+Starting Epoch 99: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[99 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.07046758; EvalErr[0]PerSample = 0.02503906; TotalTime = 11.5808s; SamplesPerSecond = 2210.5
+Finished Epoch[99 of 160]: [Training Set] TrainLossPerSample = 0.069847755; EvalErrPerSample = 0.024219999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3476
+Starting Epoch 100: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[100 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.06165749; EvalErr[0]PerSample = 0.02144531; TotalTime = 11.5168s; SamplesPerSecond = 2222.8
+Finished Epoch[100 of 160]: [Training Set] TrainLossPerSample = 0.066228345; EvalErrPerSample = 0.02266; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2882
+Starting Epoch 101: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[101 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.06483539; EvalErr[0]PerSample = 0.02242188; TotalTime = 11.5478s; SamplesPerSecond = 2216.9
+Finished Epoch[101 of 160]: [Training Set] TrainLossPerSample = 0.066535138; EvalErrPerSample = 0.022839999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.382
+Starting Epoch 102: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[102 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.06374743; EvalErr[0]PerSample = 0.02113281; TotalTime = 11.7412s; SamplesPerSecond = 2180.4
+Finished Epoch[102 of 160]: [Training Set] TrainLossPerSample = 0.06468109; EvalErrPerSample = 0.021979999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.5325
+Starting Epoch 103: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[103 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.06524000; EvalErr[0]PerSample = 0.02261719; TotalTime = 11.5928s; SamplesPerSecond = 2208.3
+Finished Epoch[103 of 160]: [Training Set] TrainLossPerSample = 0.065452188; EvalErrPerSample = 0.021879999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3876
+Starting Epoch 104: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[104 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.06336018; EvalErr[0]PerSample = 0.02117188; TotalTime = 11.5741s; SamplesPerSecond = 2211.8
+Finished Epoch[104 of 160]: [Training Set] TrainLossPerSample = 0.064649269; EvalErrPerSample = 0.021819999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3409
+Starting Epoch 105: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[105 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05928923; EvalErr[0]PerSample = 0.02039062; TotalTime = 11.4805s; SamplesPerSecond = 2229.9
+Finished Epoch[105 of 160]: [Training Set] TrainLossPerSample = 0.062256567; EvalErrPerSample = 0.021199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2864
+Starting Epoch 106: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[106 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05956097; EvalErr[0]PerSample = 0.02070312; TotalTime = 11.5877s; SamplesPerSecond = 2209.2
+Finished Epoch[106 of 160]: [Training Set] TrainLossPerSample = 0.059500545; EvalErrPerSample = 0.02076; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3573
+Starting Epoch 107: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[107 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05682479; EvalErr[0]PerSample = 0.01898438; TotalTime = 11.4687s; SamplesPerSecond = 2232.2
+Finished Epoch[107 of 160]: [Training Set] TrainLossPerSample = 0.05800005; EvalErrPerSample = 0.019579999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3035
+Starting Epoch 108: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[108 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05657757; EvalErr[0]PerSample = 0.01890625; TotalTime = 11.5761s; SamplesPerSecond = 2211.4
+Finished Epoch[108 of 160]: [Training Set] TrainLossPerSample = 0.059094436; EvalErrPerSample = 0.02004; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3386
+Starting Epoch 109: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[109 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05917750; EvalErr[0]PerSample = 0.01945313; TotalTime = 11.5089s; SamplesPerSecond = 2224.4
+Finished Epoch[109 of 160]: [Training Set] TrainLossPerSample = 0.057597388; EvalErrPerSample = 0.01898; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2734
+Starting Epoch 110: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[110 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05674717; EvalErr[0]PerSample = 0.01968750; TotalTime = 11.5447s; SamplesPerSecond = 2217.5
+Finished Epoch[110 of 160]: [Training Set] TrainLossPerSample = 0.05731637; EvalErrPerSample = 0.019819999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3039
+Starting Epoch 111: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[111 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05444533; EvalErr[0]PerSample = 0.01945313; TotalTime = 11.4688s; SamplesPerSecond = 2232.1
+Finished Epoch[111 of 160]: [Training Set] TrainLossPerSample = 0.055398725; EvalErrPerSample = 0.01956; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.244
+Starting Epoch 112: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[112 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05179993; EvalErr[0]PerSample = 0.01710938; TotalTime = 11.5138s; SamplesPerSecond = 2223.4
+Finished Epoch[112 of 160]: [Training Set] TrainLossPerSample = 0.053484067; EvalErrPerSample = 0.018139999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.273
+Starting Epoch 113: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[113 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05144035; EvalErr[0]PerSample = 0.01714844; TotalTime = 11.5189s; SamplesPerSecond = 2222.4
+Finished Epoch[113 of 160]: [Training Set] TrainLossPerSample = 0.053684823; EvalErrPerSample = 0.018139999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.298
+Starting Epoch 114: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[114 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05280342; EvalErr[0]PerSample = 0.01742188; TotalTime = 11.5954s; SamplesPerSecond = 2207.8
+Finished Epoch[114 of 160]: [Training Set] TrainLossPerSample = 0.05435824; EvalErrPerSample = 0.0184; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.4052
+Starting Epoch 115: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[115 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05222118; EvalErr[0]PerSample = 0.01757813; TotalTime = 11.6020s; SamplesPerSecond = 2206.5
+Finished Epoch[115 of 160]: [Training Set] TrainLossPerSample = 0.05275812; EvalErrPerSample = 0.0177; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3742
+Starting Epoch 116: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[116 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05418136; EvalErr[0]PerSample = 0.01832031; TotalTime = 11.5462s; SamplesPerSecond = 2217.2
+Finished Epoch[116 of 160]: [Training Set] TrainLossPerSample = 0.05442594; EvalErrPerSample = 0.018379999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3169
+Starting Epoch 117: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[117 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05089517; EvalErr[0]PerSample = 0.01753906; TotalTime = 11.4854s; SamplesPerSecond = 2228.9
+Finished Epoch[117 of 160]: [Training Set] TrainLossPerSample = 0.051556218; EvalErrPerSample = 0.018059999; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.2698
+Starting Epoch 118: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[118 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04923437; EvalErr[0]PerSample = 0.01722656; TotalTime = 11.6094s; SamplesPerSecond = 2205.1
+Finished Epoch[118 of 160]: [Training Set] TrainLossPerSample = 0.051423203; EvalErrPerSample = 0.01756; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3705
+Starting Epoch 119: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[119 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04878937; EvalErr[0]PerSample = 0.01687500; TotalTime = 11.5243s; SamplesPerSecond = 2221.4
+Finished Epoch[119 of 160]: [Training Set] TrainLossPerSample = 0.049813163; EvalErrPerSample = 0.01692; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3001
+Starting Epoch 120: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[120 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04832996; EvalErr[0]PerSample = 0.01691406; TotalTime = 11.4980s; SamplesPerSecond = 2226.5
+Finished Epoch[120 of 160]: [Training Set] TrainLossPerSample = 0.049110278; EvalErrPerSample = 0.01716; AvgLearningRatePerSample = 0.00078125001; EpochTime=22.3303
+Starting Epoch 121: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[121 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04482439; EvalErr[0]PerSample = 0.01488281; TotalTime = 11.5958s; SamplesPerSecond = 2207.7
+Finished Epoch[121 of 160]: [Training Set] TrainLossPerSample = 0.043392066; EvalErrPerSample = 0.01438; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4054
+Starting Epoch 122: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[122 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04098450; EvalErr[0]PerSample = 0.01304687; TotalTime = 11.5422s; SamplesPerSecond = 2217.9
+Finished Epoch[122 of 160]: [Training Set] TrainLossPerSample = 0.040084258; EvalErrPerSample = 0.0129; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4305
+Starting Epoch 123: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[123 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03640993; EvalErr[0]PerSample = 0.01175781; TotalTime = 11.5831s; SamplesPerSecond = 2210.1
+Finished Epoch[123 of 160]: [Training Set] TrainLossPerSample = 0.03892789; EvalErrPerSample = 0.01236; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3743
+Starting Epoch 124: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[124 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03714126; EvalErr[0]PerSample = 0.01171875; TotalTime = 11.5510s; SamplesPerSecond = 2216.3
+Finished Epoch[124 of 160]: [Training Set] TrainLossPerSample = 0.038824685; EvalErrPerSample = 0.01272; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3909
+Starting Epoch 125: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[125 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03836004; EvalErr[0]PerSample = 0.01242188; TotalTime = 11.5348s; SamplesPerSecond = 2219.4
+Finished Epoch[125 of 160]: [Training Set] TrainLossPerSample = 0.03861165; EvalErrPerSample = 0.01212; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2952
+Starting Epoch 126: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[126 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03456074; EvalErr[0]PerSample = 0.01113281; TotalTime = 11.5272s; SamplesPerSecond = 2220.8
+Finished Epoch[126 of 160]: [Training Set] TrainLossPerSample = 0.035167199; EvalErrPerSample = 0.01126; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2908
+Starting Epoch 127: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[127 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03603784; EvalErr[0]PerSample = 0.01101562; TotalTime = 11.3989s; SamplesPerSecond = 2245.8
+Finished Epoch[127 of 160]: [Training Set] TrainLossPerSample = 0.035970744; EvalErrPerSample = 0.01098; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.226
+Starting Epoch 128: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[128 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03553395; EvalErr[0]PerSample = 0.01175781; TotalTime = 11.5519s; SamplesPerSecond = 2216.1
+Finished Epoch[128 of 160]: [Training Set] TrainLossPerSample = 0.035689104; EvalErrPerSample = 0.01156; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3882
+Starting Epoch 129: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[129 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03523235; EvalErr[0]PerSample = 0.01109375; TotalTime = 11.5823s; SamplesPerSecond = 2210.3
+Finished Epoch[129 of 160]: [Training Set] TrainLossPerSample = 0.036777027; EvalErrPerSample = 0.01176; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3416
+Starting Epoch 130: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[130 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03563953; EvalErr[0]PerSample = 0.01058594; TotalTime = 11.4850s; SamplesPerSecond = 2229.0
+Finished Epoch[130 of 160]: [Training Set] TrainLossPerSample = 0.034818619; EvalErrPerSample = 0.010559999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2438
+Starting Epoch 131: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[131 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03574152; EvalErr[0]PerSample = 0.01187500; TotalTime = 11.4836s; SamplesPerSecond = 2229.3
+Finished Epoch[131 of 160]: [Training Set] TrainLossPerSample = 0.035892151; EvalErrPerSample = 0.01174; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2433
+Starting Epoch 132: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[132 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03485843; EvalErr[0]PerSample = 0.01152344; TotalTime = 11.4668s; SamplesPerSecond = 2232.5
+Finished Epoch[132 of 160]: [Training Set] TrainLossPerSample = 0.035623327; EvalErrPerSample = 0.01148; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2302
+Starting Epoch 133: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[133 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03582706; EvalErr[0]PerSample = 0.01058594; TotalTime = 11.4449s; SamplesPerSecond = 2236.8
+Finished Epoch[133 of 160]: [Training Set] TrainLossPerSample = 0.035650004; EvalErrPerSample = 0.01076; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2876
+Starting Epoch 134: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[134 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03308557; EvalErr[0]PerSample = 0.01031250; TotalTime = 11.5628s; SamplesPerSecond = 2214.0
+Finished Epoch[134 of 160]: [Training Set] TrainLossPerSample = 0.034471795; EvalErrPerSample = 0.010679999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3507
+Starting Epoch 135: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[135 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03313774; EvalErr[0]PerSample = 0.01011719; TotalTime = 11.5699s; SamplesPerSecond = 2212.6
+Finished Epoch[135 of 160]: [Training Set] TrainLossPerSample = 0.03418766; EvalErrPerSample = 0.01052; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.356
+Starting Epoch 136: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[136 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03477723; EvalErr[0]PerSample = 0.01121094; TotalTime = 11.5129s; SamplesPerSecond = 2223.6
+Finished Epoch[136 of 160]: [Training Set] TrainLossPerSample = 0.0342177; EvalErrPerSample = 0.01064; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2738
+Starting Epoch 137: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[137 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03409709; EvalErr[0]PerSample = 0.01007813; TotalTime = 11.4740s; SamplesPerSecond = 2231.1
+Finished Epoch[137 of 160]: [Training Set] TrainLossPerSample = 0.034409851; EvalErrPerSample = 0.01076; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2606
+Starting Epoch 138: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[138 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03408328; EvalErr[0]PerSample = 0.01019531; TotalTime = 11.6197s; SamplesPerSecond = 2203.1
+Finished Epoch[138 of 160]: [Training Set] TrainLossPerSample = 0.03398142; EvalErrPerSample = 0.01034; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4016
+Starting Epoch 139: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[139 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03391739; EvalErr[0]PerSample = 0.01003906; TotalTime = 11.5855s; SamplesPerSecond = 2209.7
+Finished Epoch[139 of 160]: [Training Set] TrainLossPerSample = 0.034369528; EvalErrPerSample = 0.0105; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4288
+Starting Epoch 140: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[140 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03432070; EvalErr[0]PerSample = 0.01093750; TotalTime = 11.5739s; SamplesPerSecond = 2211.9
+Finished Epoch[140 of 160]: [Training Set] TrainLossPerSample = 0.034433346; EvalErrPerSample = 0.01072; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3479
+Starting Epoch 141: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[141 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03326862; EvalErr[0]PerSample = 0.01023438; TotalTime = 11.5483s; SamplesPerSecond = 2216.8
+Finished Epoch[141 of 160]: [Training Set] TrainLossPerSample = 0.03355087; EvalErrPerSample = 0.01038; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.326
+Starting Epoch 142: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[142 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03197437; EvalErr[0]PerSample = 0.00988281; TotalTime = 11.4156s; SamplesPerSecond = 2242.6
+Finished Epoch[142 of 160]: [Training Set] TrainLossPerSample = 0.032701213; EvalErrPerSample = 0.0098000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.1816
+Starting Epoch 143: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[143 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03477237; EvalErr[0]PerSample = 0.01089844; TotalTime = 11.5553s; SamplesPerSecond = 2215.4
+Finished Epoch[143 of 160]: [Training Set] TrainLossPerSample = 0.034088843; EvalErrPerSample = 0.01066; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3259
+Starting Epoch 144: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[144 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03017725; EvalErr[0]PerSample = 0.00953125; TotalTime = 11.5897s; SamplesPerSecond = 2208.9
+Finished Epoch[144 of 160]: [Training Set] TrainLossPerSample = 0.031698398; EvalErrPerSample = 0.0099999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.5923
+Starting Epoch 145: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[145 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03448967; EvalErr[0]PerSample = 0.01074219; TotalTime = 11.4235s; SamplesPerSecond = 2241.0
+Finished Epoch[145 of 160]: [Training Set] TrainLossPerSample = 0.033083703; EvalErrPerSample = 0.01002; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.1806
+Starting Epoch 146: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[146 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03416681; EvalErr[0]PerSample = 0.01136719; TotalTime = 11.3855s; SamplesPerSecond = 2248.5
+Finished Epoch[146 of 160]: [Training Set] TrainLossPerSample = 0.034353275; EvalErrPerSample = 0.010799999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.1946
+Starting Epoch 147: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[147 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03433434; EvalErr[0]PerSample = 0.01101562; TotalTime = 11.5644s; SamplesPerSecond = 2213.7
+Finished Epoch[147 of 160]: [Training Set] TrainLossPerSample = 0.034649722; EvalErrPerSample = 0.011279999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3273
+Starting Epoch 148: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[148 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03219655; EvalErr[0]PerSample = 0.00988281; TotalTime = 11.5637s; SamplesPerSecond = 2213.8
+Finished Epoch[148 of 160]: [Training Set] TrainLossPerSample = 0.031285692; EvalErrPerSample = 0.0094999997; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3287
+Starting Epoch 149: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[149 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03190110; EvalErr[0]PerSample = 0.01046875; TotalTime = 11.4602s; SamplesPerSecond = 2233.8
+Finished Epoch[149 of 160]: [Training Set] TrainLossPerSample = 0.031930499; EvalErrPerSample = 0.0101; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2852
+Starting Epoch 150: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[150 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03089277; EvalErr[0]PerSample = 0.00921875; TotalTime = 11.5981s; SamplesPerSecond = 2207.3
+Finished Epoch[150 of 160]: [Training Set] TrainLossPerSample = 0.031741869; EvalErrPerSample = 0.00954; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4059
+Starting Epoch 151: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[151 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03341821; EvalErr[0]PerSample = 0.01105469; TotalTime = 11.5961s; SamplesPerSecond = 2207.6
+Finished Epoch[151 of 160]: [Training Set] TrainLossPerSample = 0.032794204; EvalErrPerSample = 0.01036; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.435
+Starting Epoch 152: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[152 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03127065; EvalErr[0]PerSample = 0.00937500; TotalTime = 11.5746s; SamplesPerSecond = 2211.7
+Finished Epoch[152 of 160]: [Training Set] TrainLossPerSample = 0.032228082; EvalErrPerSample = 0.0098599996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3883
+Starting Epoch 153: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[153 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03200430; EvalErr[0]PerSample = 0.01007813; TotalTime = 11.5357s; SamplesPerSecond = 2219.2
+Finished Epoch[153 of 160]: [Training Set] TrainLossPerSample = 0.031472486; EvalErrPerSample = 0.0098999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2992
+Starting Epoch 154: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[154 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03112781; EvalErr[0]PerSample = 0.00890625; TotalTime = 11.4339s; SamplesPerSecond = 2239.0
+Finished Epoch[154 of 160]: [Training Set] TrainLossPerSample = 0.030955072; EvalErrPerSample = 0.00954; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.2657
+Starting Epoch 155: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[155 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03170617; EvalErr[0]PerSample = 0.00941406; TotalTime = 11.5475s; SamplesPerSecond = 2216.9
+Finished Epoch[155 of 160]: [Training Set] TrainLossPerSample = 0.031180983; EvalErrPerSample = 0.0093200002; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3148
+Starting Epoch 156: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[156 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02884187; EvalErr[0]PerSample = 0.00929687; TotalTime = 11.4932s; SamplesPerSecond = 2227.4
+Finished Epoch[156 of 160]: [Training Set] TrainLossPerSample = 0.030790305; EvalErrPerSample = 0.0097199995; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3262
+Starting Epoch 157: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[157 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03166680; EvalErr[0]PerSample = 0.00992188; TotalTime = 11.6570s; SamplesPerSecond = 2196.1
+Finished Epoch[157 of 160]: [Training Set] TrainLossPerSample = 0.031061091; EvalErrPerSample = 0.0094599994; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.4556
+Starting Epoch 158: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[158 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03136421; EvalErr[0]PerSample = 0.00906250; TotalTime = 11.5571s; SamplesPerSecond = 2215.1
+Finished Epoch[158 of 160]: [Training Set] TrainLossPerSample = 0.029945096; EvalErrPerSample = 0.0084999995; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3185
+Starting Epoch 159: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[159 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03113249; EvalErr[0]PerSample = 0.00976563; TotalTime = 11.5817s; SamplesPerSecond = 2210.4
+Finished Epoch[159 of 160]: [Training Set] TrainLossPerSample = 0.03114122; EvalErrPerSample = 0.0093599996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.343
+Starting Epoch 160: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[160 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03009957; EvalErr[0]PerSample = 0.00886719; TotalTime = 11.5242s; SamplesPerSecond = 2221.4
+Finished Epoch[160 of 160]: [Training Set] TrainLossPerSample = 0.03085361; EvalErrPerSample = 0.00942; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=22.3018
+CNTKCommandTrainEnd: Train
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+
+Post-processing network...
+
+3 roots:
+	OutputNodes.z = Plus
+	Err = ErrorPrediction
+	CE = CrossEntropyWithSoftmax
+FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
+FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
+FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
+
+
+Validating for node OutputNodes.z. 185 nodes to process in pass 1.
+
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+
+Validating for node OutputNodes.z. 77 nodes to process in pass 2.
+
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+
+Validating for node OutputNodes.z, final verification.
+
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1])WARNING: trying to use cuDNN on unsupported platform. It is safe to ignore the warning if it's produced during model editing command.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+
+107 out of 185 nodes do not share the minibatch layout with the input data.
+
+
+Validating for node Err. 187 nodes to process in pass 1.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node Err. 78 nodes to process in pass 2.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node Err, final verification.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+108 out of 187 nodes do not share the minibatch layout with the input data.
+
+
+Validating for node CE. 187 nodes to process in pass 1.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node CE. 78 nodes to process in pass 2.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node CE, final verification.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+108 out of 187 nodes do not share the minibatch layout with the input data.
+
+Post-processing network complete.
+
+Post-processing network...
+
+3 roots:
+	OutputNodes.z = Plus
+	Err = ErrorPrediction
+	CE = CrossEntropyWithSoftmax
+FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
+FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
+FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
+
+
+Validating for node OutputNodes.z. 185 nodes to process in pass 1.
+
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+
+Validating for node OutputNodes.z. 77 nodes to process in pass 2.
+
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+
+Validating for node OutputNodes.z, final verification.
+
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+
+107 out of 185 nodes do not share the minibatch layout with the input data.
+
+
+Validating for node Err. 187 nodes to process in pass 1.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node Err. 78 nodes to process in pass 2.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node Err, final verification.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> Err = ErrorPrediction(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+108 out of 187 nodes do not share the minibatch layout with the input data.
+
+
+Validating for node CE. 187 nodes to process in pass 1.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node CE. 78 nodes to process in pass 2.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+Validating for node CE, final verification.
+
+Validating --> labels = InputValue -> [10 [10], MBSize 0]
+Validating --> OutputNodes.W = LearnableParameter -> [10 [10], 64]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 [64], 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 [64], 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 [32], 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 [32], 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 [16], 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 [16], 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 [16], 27]
+Validating --> features = InputValue -> [3072 [32 x 32 x 3], MBSize 0]
+Validating --> conv1.c.c = Convolution(conv1.c.W[16, 27], features[3072 [32 x 32 x 3] {W=32, H=3, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> conv1.c.y = BatchNormalization(conv1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.c.sc[16, 1], conv1.c.b[16, 1], conv1.c.m[16, 1], conv1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> conv1.y = RectifiedLinear(conv1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.c = Convolution(rn1_1.c1.c.W[16, 144], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c1.c.y = BatchNormalization(rn1_1.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c1.c.sc[16, 1], rn1_1.c1.c.b[16, 1], rn1_1.c1.c.m[16, 1], rn1_1.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.c = Convolution(rn1_1.c2.W[16, 144], rn1_1.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_1.c2.y = BatchNormalization(rn1_1.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.c2.sc[16, 1], rn1_1.c2.b[16, 1], rn1_1.c2.m[16, 1], rn1_1.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.p = Plus(rn1_1.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], conv1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.c = Convolution(rn1_2.c1.c.W[16, 144], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c1.c.y = BatchNormalization(rn1_2.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c1.c.sc[16, 1], rn1_2.c1.c.b[16, 1], rn1_2.c1.c.m[16, 1], rn1_2.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.c = Convolution(rn1_2.c2.W[16, 144], rn1_2.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_2.c2.y = BatchNormalization(rn1_2.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.c2.sc[16, 1], rn1_2.c2.b[16, 1], rn1_2.c2.m[16, 1], rn1_2.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.p = Plus(rn1_2.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.c = Convolution(rn1_3.c1.c.W[16, 144], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.c.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c1.c.y = BatchNormalization(rn1_3.c1.c.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c1.c.sc[16, 1], rn1_3.c1.c.b[16, 1], rn1_3.c1.c.m[16, 1], rn1_3.c1.c.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.c = Convolution(rn1_3.c2.W[16, 144], rn1_3.c1.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn1_3.c2.y = BatchNormalization(rn1_3.c2.c[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_3.c2.sc[16, 1], rn1_3.c2.b[16, 1], rn1_3.c2.m[16, 1], rn1_3.c2.isd[16, 1]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.p = Plus(rn1_3.c2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0], rn1_2.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [16384 [32 x 32 x 16], MBSize 0]
+Validating --> rn2_1.c1.c.c = Convolution(rn2_1.c1.c.W[32, 144], rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_1.c1.c.y = BatchNormalization(rn2_1.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.c1.c.sc[32, 1], rn2_1.c1.c.b[32, 1], rn2_1.c1.c.m[32, 1], rn2_1.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.c2.c = Convolution(rn2_1.c2.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c2.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c2.y = BatchNormalization(rn2_1.c2.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c2.sc[16, 1], rn2_1.c2.b[16, 1], rn2_1.c2.m[16, 1], rn2_1.c2.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.pool = MaxPooling(rn1_3.y[16384 [32 x 32 x 16] {W=32, H=16, C=32}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.p = Plus(rn2_1.c2.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.pool[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.W = LearnableParameter -> [16 [16], 288]
+Validating --> rn2_1.c3.c = Convolution(rn2_1.c3.W[16, 288], rn2_1.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.c3.sc = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.b = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.m = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.isd = LearnableParameter -> [16 [16], 1]
+Validating --> rn2_1.c3.y = BatchNormalization(rn2_1.c3.c[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.sc[16, 1], rn2_1.c3.b[16, 1], rn2_1.c3.m[16, 1], rn2_1.c3.isd[16, 1]) -> [4096 [16 x 16 x 16], MBSize 0]
+Validating --> rn2_1.r = RowStack(rn2_1.p[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0], rn2_1.c3.y[4096 [16 x 16 x 16] {W=16, H=16, C=16}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.r[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.c = Convolution(rn2_2.c1.c.W[32, 288], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c1.c.y = BatchNormalization(rn2_2.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c1.c.sc[32, 1], rn2_2.c1.c.b[32, 1], rn2_2.c1.c.m[32, 1], rn2_2.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.c = Convolution(rn2_2.c2.W[32, 288], rn2_2.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_2.c2.y = BatchNormalization(rn2_2.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.c2.sc[32, 1], rn2_2.c2.b[32, 1], rn2_2.c2.m[32, 1], rn2_2.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.p = Plus(rn2_2.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.c = Convolution(rn2_3.c1.c.W[32, 288], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.c.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c1.c.y = BatchNormalization(rn2_3.c1.c.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c1.c.sc[32, 1], rn2_3.c1.c.b[32, 1], rn2_3.c1.c.m[32, 1], rn2_3.c1.c.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.c = Convolution(rn2_3.c2.W[32, 288], rn2_3.c1.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn2_3.c2.y = BatchNormalization(rn2_3.c2.c[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_3.c2.sc[32, 1], rn2_3.c2.b[32, 1], rn2_3.c2.m[32, 1], rn2_3.c2.isd[32, 1]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.p = Plus(rn2_3.c2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0], rn2_2.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [8192 [16 x 16 x 32], MBSize 0]
+Validating --> rn3_1.c1.c.c = Convolution(rn3_1.c1.c.W[64, 288], rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_1.c1.c.y = BatchNormalization(rn3_1.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.c1.c.sc[64, 1], rn3_1.c1.c.b[64, 1], rn3_1.c1.c.m[64, 1], rn3_1.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.c2.c = Convolution(rn3_1.c2.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c2.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c2.y = BatchNormalization(rn3_1.c2.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c2.sc[32, 1], rn3_1.c2.b[32, 1], rn3_1.c2.m[32, 1], rn3_1.c2.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.pool = MaxPooling(rn2_3.y[8192 [16 x 16 x 32] {W=16, H=32, C=16}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.p = Plus(rn3_1.c2.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.pool[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.W = LearnableParameter -> [32 [32], 576]
+Validating --> rn3_1.c3.c = Convolution(rn3_1.c3.W[32, 576], rn3_1.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.c3.sc = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.b = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.m = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.isd = LearnableParameter -> [32 [32], 1]
+Validating --> rn3_1.c3.y = BatchNormalization(rn3_1.c3.c[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.sc[32, 1], rn3_1.c3.b[32, 1], rn3_1.c3.m[32, 1], rn3_1.c3.isd[32, 1]) -> [2048 [8 x 8 x 32], MBSize 0]
+Validating --> rn3_1.r = RowStack(rn3_1.p[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0], rn3_1.c3.y[2048 [8 x 8 x 32] {W=8, H=32, C=8}, MBSize 0])WARNING: RowStack operation cannot inherit image size information from its child. Image size info is lost.
+ -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.r[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.c = Convolution(rn3_2.c1.c.W[64, 576], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c1.c.y = BatchNormalization(rn3_2.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c1.c.sc[64, 1], rn3_2.c1.c.b[64, 1], rn3_2.c1.c.m[64, 1], rn3_2.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.c = Convolution(rn3_2.c2.W[64, 576], rn3_2.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_2.c2.y = BatchNormalization(rn3_2.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.c2.sc[64, 1], rn3_2.c2.b[64, 1], rn3_2.c2.m[64, 1], rn3_2.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.p = Plus(rn3_2.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.c = Convolution(rn3_3.c1.c.W[64, 576], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.c.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c1.c.y = BatchNormalization(rn3_3.c1.c.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c1.c.sc[64, 1], rn3_3.c1.c.b[64, 1], rn3_3.c1.c.m[64, 1], rn3_3.c1.c.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.c = Convolution(rn3_3.c2.W[64, 576], rn3_3.c1.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.c2.sc = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.b = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.m = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.isd = LearnableParameter -> [64 [64], 1]
+Validating --> rn3_3.c2.y = BatchNormalization(rn3_3.c2.c[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_3.c2.sc[64, 1], rn3_3.c2.b[64, 1], rn3_3.c2.m[64, 1], rn3_3.c2.isd[64, 1]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.p = Plus(rn3_3.c2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0], rn3_2.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [4096 [8 x 8 x 64], MBSize 0]
+Validating --> pool = AveragePooling(rn3_3.y[4096 [8 x 8 x 64] {W=8, H=64, C=8}, MBSize 0]) -> [64 [1 x 1 x 64], MBSize 0]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10, 64], pool[64 [1 x 1 x 64], MBSize 0]) -> [10 [10], MBSize 0]
+Validating --> OutputNodes.b = LearnableParameter -> [10 [10], 1]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10, MBSize 0], OutputNodes.b[10, 1]) -> [10 [10], MBSize 0]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10, MBSize 0], OutputNodes.z[10, MBSize 0]) -> [1 [1], 1]
+
+108 out of 187 nodes do not share the minibatch layout with the input data.
+
+Post-processing network complete.
+evalNodeNames are not specified, using all the default evalnodes and training criterion nodes.
+
+
+Allocating matrices for forward and/or backward propagation.
+Minibatch[1-20]: Samples Seen = 10000    Err: ErrorPrediction/Sample = 0.0819    CE: CrossEntropyWithSoftmax/Sample = 0.35141698    
+Final Results: Minibatch[1-20]: Samples Seen = 10000    Err: ErrorPrediction/Sample = 0.0819    CE: CrossEntropyWithSoftmax/Sample = 0.35141698    Perplexity = 1.4210798    
+__COMPLETED__
diff --git a/Examples/Image/Miscellaneous/CIFAR-10/Output/04_ResNet_56_Train_AddBNEval_Test.log.160 b/Examples/Image/Miscellaneous/CIFAR-10/Output/04_ResNet_56_Train_AddBNEval_Test.log.160
index b309edeed..d1de5c398 100644
--- a/Examples/Image/Miscellaneous/CIFAR-10/Output/04_ResNet_56_Train_AddBNEval_Test.log.160
+++ b/Examples/Image/Miscellaneous/CIFAR-10/Output/04_ResNet_56_Train_AddBNEval_Test.log.160
@@ -1,9902 +1,9902 @@
--------------------------------------------------------------------
-Build info: 
-
-		Built time: Feb 26 2016 16:36:12
-		Last modified date: Thu Feb 25 12:56:12 2016
-		Build type: Release
-		Build target: GPU
-		With 1bit-SGD: no
-		CUDA_PATH: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5
-		CUB_PATH: C:\src\cub
-		CUDNN_PATH: C:\NVIDIA\cudnn-4.0\cuda
-		Build Branch: 
-		Build SHA1:  (modified)
-		Built by alexeyk on z840-01
-		Build Path: C:\src\cntk\Source\CNTK\
--------------------------------------------------------------------
-running on z840-01 at 2016/02/26 17:43:42
-command line: 
-C:\src\cntk\x64\Release\CNTK.exe configFile=04_ResNet_56.cntk 
-
->>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>
-RootDir = "."
-ConfigDir = "$RootDir$"
-DataDir = "$RootDir$"
-OutputDir = "$RootDir$/Output"
-ModelDir = "$OutputDir$/Models"
-ndlMacros="$ConfigDir$/Macros.ndl"
-precision="float"
-deviceId="Auto"
-prefetch="true"
-parallelTrain="false"
-command=Train:AddBNEval:Test
-stderr="$OutputDir$/04_ResNet_56"
-traceLevel=1
-numMBsToShowResult=200
-Proj16to32Filename = "$ConfigDir$/16to32.txt"
-Proj32to64Filename = "$ConfigDir$/32to64.txt"
-Train=[
-    action="train"
-    modelPath="$ModelDir$/04_ResNet_56"
-     NDLNetworkBuilder=[
-        networkDescription="$ConfigDir$/04_ResNet_56.ndl"
-    ]
-    SGD=[
-        epochSize=0
-        minibatchSize=128
-        learningRatesPerMB=0.1*1:1.0*80:0.1*40:0.01
-        momentumPerMB=0.9
-        maxEpochs=160
-        L2RegWeight=0.0001
-        dropoutRate=0
-        ParallelTrain=[
-            parallelizationMethod="DataParallelSGD"
-            distributedMBReading="true"
-            parallelizationStartEpoch=1
-            DataParallelSGD=[
-                gradientBits=32
-            ]
-        ]
-    ]
-    reader=[
-        readerType="ImageReader"
-        file="$DataDir$/train_map.txt"
-        randomize="Auto"
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType="Random"
-            cropRatio=0.8
-            jitterType="UniRatio"
-            interpolations="Linear"
-            meanFile="$ConfigDir$/CIFAR-10_mean.xml"
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-AddBNEval=[    
-    action="edit"
-    CurModel="$ModelDir$/04_ResNet_56"
-    NewModel="$ModelDir$/04_ResNet_56.Eval"
-    editPath="$ConfigDir$/03_ResNet.mel"
-]
-Test=[
-    action="test"
-    modelPath="$ModelDir$/04_ResNet_56.Eval"
-    minibatchSize=512
-    reader=[
-        readerType="ImageReader"
-        file="$DataDir$/test_map.txt"
-        randomize="None"
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType="Center"
-            cropRatio=1
-            jitterType="UniRatio"
-            interpolations="Linear"
-            meanFile="$ConfigDir$/CIFAR-10_mean.xml"
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-
-<<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<
-
->>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
-RootDir = "."
-ConfigDir = "."
-DataDir = "."
-OutputDir = "./Output"
-ModelDir = "./Output/Models"
-ndlMacros="./Macros.ndl"
-precision="float"
-deviceId="Auto"
-prefetch="true"
-parallelTrain="false"
-command=Train:AddBNEval:Test
-stderr="./Output/04_ResNet_56"
-traceLevel=1
-numMBsToShowResult=200
-Proj16to32Filename = "./16to32.txt"
-Proj32to64Filename = "./32to64.txt"
-Train=[
-    action="train"
-    modelPath="./Output/Models/04_ResNet_56"
-     NDLNetworkBuilder=[
-        networkDescription="./04_ResNet_56.ndl"
-    ]
-    SGD=[
-        epochSize=0
-        minibatchSize=128
-        learningRatesPerMB=0.1*1:1.0*80:0.1*40:0.01
-        momentumPerMB=0.9
-        maxEpochs=160
-        L2RegWeight=0.0001
-        dropoutRate=0
-        ParallelTrain=[
-            parallelizationMethod="DataParallelSGD"
-            distributedMBReading="true"
-            parallelizationStartEpoch=1
-            DataParallelSGD=[
-                gradientBits=32
-            ]
-        ]
-    ]
-    reader=[
-        readerType="ImageReader"
-        file="./train_map.txt"
-        randomize="Auto"
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType="Random"
-            cropRatio=0.8
-            jitterType="UniRatio"
-            interpolations="Linear"
-            meanFile="./CIFAR-10_mean.xml"
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-AddBNEval=[    
-    action="edit"
-    CurModel="./Output/Models/04_ResNet_56"
-    NewModel="./Output/Models/04_ResNet_56.Eval"
-    editPath="./03_ResNet.mel"
-]
-Test=[
-    action="test"
-    modelPath="./Output/Models/04_ResNet_56.Eval"
-    minibatchSize=512
-    reader=[
-        readerType="ImageReader"
-        file="./test_map.txt"
-        randomize="None"
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType="Center"
-            cropRatio=1
-            jitterType="UniRatio"
-            interpolations="Linear"
-            meanFile="./CIFAR-10_mean.xml"
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-
-<<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
-
->>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
-configparameters: 04_ResNet_56.cntk:AddBNEval=[    
-    action="edit"
-    CurModel="./Output/Models/04_ResNet_56"
-    NewModel="./Output/Models/04_ResNet_56.Eval"
-    editPath="./03_ResNet.mel"
-]
-
-configparameters: 04_ResNet_56.cntk:command=Train:AddBNEval:Test
-configparameters: 04_ResNet_56.cntk:ConfigDir=.
-configparameters: 04_ResNet_56.cntk:DataDir=.
-configparameters: 04_ResNet_56.cntk:deviceId=Auto
-configparameters: 04_ResNet_56.cntk:ModelDir=./Output/Models
-configparameters: 04_ResNet_56.cntk:ndlMacros=./Macros.ndl
-configparameters: 04_ResNet_56.cntk:numMBsToShowResult=200
-configparameters: 04_ResNet_56.cntk:OutputDir=./Output
-configparameters: 04_ResNet_56.cntk:parallelTrain=false
-configparameters: 04_ResNet_56.cntk:precision=float
-configparameters: 04_ResNet_56.cntk:prefetch=true
-configparameters: 04_ResNet_56.cntk:Proj16to32Filename=./16to32.txt
-configparameters: 04_ResNet_56.cntk:Proj32to64Filename=./32to64.txt
-configparameters: 04_ResNet_56.cntk:RootDir=.
-configparameters: 04_ResNet_56.cntk:stderr=./Output/04_ResNet_56
-configparameters: 04_ResNet_56.cntk:Test=[
-    action="test"
-    modelPath="./Output/Models/04_ResNet_56.Eval"
-    minibatchSize=512
-    reader=[
-        readerType="ImageReader"
-        file="./test_map.txt"
-        randomize="None"
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType="Center"
-            cropRatio=1
-            jitterType="UniRatio"
-            interpolations="Linear"
-            meanFile="./CIFAR-10_mean.xml"
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-
-configparameters: 04_ResNet_56.cntk:traceLevel=1
-configparameters: 04_ResNet_56.cntk:Train=[
-    action="train"
-    modelPath="./Output/Models/04_ResNet_56"
-     NDLNetworkBuilder=[
-        networkDescription="./04_ResNet_56.ndl"
-    ]
-    SGD=[
-        epochSize=0
-        minibatchSize=128
-        learningRatesPerMB=0.1*1:1.0*80:0.1*40:0.01
-        momentumPerMB=0.9
-        maxEpochs=160
-        L2RegWeight=0.0001
-        dropoutRate=0
-        ParallelTrain=[
-            parallelizationMethod="DataParallelSGD"
-            distributedMBReading="true"
-            parallelizationStartEpoch=1
-            DataParallelSGD=[
-                gradientBits=32
-            ]
-        ]
-    ]
-    reader=[
-        readerType="ImageReader"
-        file="./train_map.txt"
-        randomize="Auto"
-        features=[
-            width=32
-            height=32
-            channels=3
-            cropType="Random"
-            cropRatio=0.8
-            jitterType="UniRatio"
-            interpolations="Linear"
-            meanFile="./CIFAR-10_mean.xml"
-        ]
-        labels=[
-            labelDim=10
-        ]
-    ]    
-]
-
-<<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
-command: Train AddBNEval Test 
-precision = float
-CNTKModelPath: ./Output/Models/04_ResNet_56
-CNTKCommandTrainInfo: Train : 160
-CNTKCommandTrainInfo: CNTKNoMoreCommands_Total : 160
-CNTKCommandTrainBegin: Train
-LockDevice: Locked GPU 0 to test availability.
-LockDevice: Unlocked GPU 0 after testing.
-LockDevice: Locked GPU 1 to test availability.
-LockDevice: Unlocked GPU 1 after testing.
-LockDevice: Locked GPU 2 to test availability.
-LockDevice: Unlocked GPU 2 after testing.
-LockDevice: Locked GPU 0 for exclusive use.
-NDLBuilder Using GPU 0
-Microsoft::MSR::CNTK::GPUMatrix<ElemType>::SetGaussianRandomValue (GPU): creating curand object with seed 1, sizeof(ElemType)==4
-
-Post-processing network...
-
-3 roots:
-	CE = CrossEntropyWithSoftmax
-	Err = ErrorPrediction
-	OutputNodes.z = Plus
-FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
-FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
-FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
-
-
-Validating network. 949 nodes to process in pass 1.
-
-Validating --> labels = InputValue -> [10 x *]
-Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
-Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
-Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
-Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
-Validating --> features = InputValue -> [32 x 32 x 3 x *]
-Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
-Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
-Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
-Validating --> OutputNodes.b = LearnableParameter -> [10]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-
-Validating network. 390 nodes to process in pass 2.
-
-Validating --> labels = InputValue -> [10 x *]
-Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
-Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
-Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
-Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
-Validating --> features = InputValue -> [32 x 32 x 3 x *]
-Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
-Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
-Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
-Validating --> OutputNodes.b = LearnableParameter -> [10]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-
-Validating network, final pass.
-
-Validating --> labels = InputValue -> [10 x *]
-Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
-Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
-Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
-Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
-Validating --> features = InputValue -> [32 x 32 x 3 x *]
-Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
-Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
-Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
-Validating --> OutputNodes.b = LearnableParameter -> [10]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-
-559 out of 949 nodes do not share the minibatch layout with the input data.
-
-Post-processing network complete.
-
-SGD using GPU 0.
-
-Training criterion node(s):
-	CE = CrossEntropyWithSoftmax
-
-Evaluation criterion node(s):
-	Err = ErrorPrediction
-
-
-Allocating matrices for forward and/or backward propagation.
-No PreCompute nodes found, skipping PreCompute step
-Starting Epoch 1: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 1 of 160]-Minibatch[   1- 200]: SamplesSeen = 25600; TrainLossPerSample =  2.08201447; EvalErr[0]PerSample = 0.79109375; TotalTime = 37.3101s; SamplesPerSecond = 686.1
-Finished Epoch[ 1 of 160]: [Training Set] TrainLossPerSample = 1.9284772; EvalErrPerSample = 0.73097998; AvgLearningRatePerSample = 0.00078125001; EpochTime=69.7251
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.1'
-Starting Epoch 2: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 2 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  1.80199310; EvalErr[0]PerSample = 0.68078125; TotalTime = 34.7295s; SamplesPerSecond = 737.1
-Finished Epoch[ 2 of 160]: [Training Set] TrainLossPerSample = 1.6386911; EvalErrPerSample = 0.61193997; AvgLearningRatePerSample = 0.0078125; EpochTime=67.9925
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.2'
-Starting Epoch 3: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 3 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  1.25481010; EvalErr[0]PerSample = 0.45394531; TotalTime = 34.5533s; SamplesPerSecond = 740.9
-Finished Epoch[ 3 of 160]: [Training Set] TrainLossPerSample = 1.1870104; EvalErrPerSample = 0.42773998; AvgLearningRatePerSample = 0.0078125; EpochTime=67.6327
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.3'
-Starting Epoch 4: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 4 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  1.00952705; EvalErr[0]PerSample = 0.36304687; TotalTime = 34.5927s; SamplesPerSecond = 740.0
-Finished Epoch[ 4 of 160]: [Training Set] TrainLossPerSample = 0.97715324; EvalErrPerSample = 0.34739998; AvgLearningRatePerSample = 0.0078125; EpochTime=67.5202
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.4'
-Starting Epoch 5: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 5 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.84818413; EvalErr[0]PerSample = 0.29589844; TotalTime = 34.3363s; SamplesPerSecond = 745.6
-Finished Epoch[ 5 of 160]: [Training Set] TrainLossPerSample = 0.83634055; EvalErrPerSample = 0.29328001; AvgLearningRatePerSample = 0.0078125; EpochTime=67.0961
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.5'
-Starting Epoch 6: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 6 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.74924049; EvalErr[0]PerSample = 0.26066406; TotalTime = 34.2546s; SamplesPerSecond = 747.3
-Finished Epoch[ 6 of 160]: [Training Set] TrainLossPerSample = 0.73574179; EvalErrPerSample = 0.25577998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9549
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.6'
-Starting Epoch 7: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 7 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.66103149; EvalErr[0]PerSample = 0.22968750; TotalTime = 34.2366s; SamplesPerSecond = 747.7
-Finished Epoch[ 7 of 160]: [Training Set] TrainLossPerSample = 0.65649849; EvalErrPerSample = 0.22679999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8964
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.7'
-Starting Epoch 8: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 8 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.59962826; EvalErr[0]PerSample = 0.20585938; TotalTime = 34.2398s; SamplesPerSecond = 747.7
-Finished Epoch[ 8 of 160]: [Training Set] TrainLossPerSample = 0.59828544; EvalErrPerSample = 0.20622; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8987
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.8'
-Starting Epoch 9: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[ 9 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.56160091; EvalErr[0]PerSample = 0.19425781; TotalTime = 34.2457s; SamplesPerSecond = 747.5
-Finished Epoch[ 9 of 160]: [Training Set] TrainLossPerSample = 0.56435287; EvalErrPerSample = 0.19484; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9167
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.9'
-Starting Epoch 10: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[10 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.52267384; EvalErr[0]PerSample = 0.18140625; TotalTime = 34.2539s; SamplesPerSecond = 747.4
-Finished Epoch[10 of 160]: [Training Set] TrainLossPerSample = 0.517959; EvalErrPerSample = 0.17936; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9219
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.10'
-Starting Epoch 11: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[11 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.49345413; EvalErr[0]PerSample = 0.17046875; TotalTime = 34.2380s; SamplesPerSecond = 747.7
-Finished Epoch[11 of 160]: [Training Set] TrainLossPerSample = 0.48473847; EvalErrPerSample = 0.16725999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9081
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.11'
-Starting Epoch 12: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[12 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.46631210; EvalErr[0]PerSample = 0.16121094; TotalTime = 34.2568s; SamplesPerSecond = 747.3
-Finished Epoch[12 of 160]: [Training Set] TrainLossPerSample = 0.46427867; EvalErrPerSample = 0.16023999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.989
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.12'
-Starting Epoch 13: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[13 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.44940582; EvalErr[0]PerSample = 0.15414063; TotalTime = 34.2965s; SamplesPerSecond = 746.4
-Finished Epoch[13 of 160]: [Training Set] TrainLossPerSample = 0.44242746; EvalErrPerSample = 0.15223999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9538
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.13'
-Starting Epoch 14: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[14 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.41690651; EvalErr[0]PerSample = 0.14292969; TotalTime = 34.2325s; SamplesPerSecond = 747.8
-Finished Epoch[14 of 160]: [Training Set] TrainLossPerSample = 0.41924006; EvalErrPerSample = 0.1444; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8936
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.14'
-Starting Epoch 15: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[15 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.39440144; EvalErr[0]PerSample = 0.13656250; TotalTime = 34.2478s; SamplesPerSecond = 747.5
-Finished Epoch[15 of 160]: [Training Set] TrainLossPerSample = 0.39710063; EvalErrPerSample = 0.13676; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9095
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.15'
-Starting Epoch 16: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[16 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.38027695; EvalErr[0]PerSample = 0.13261719; TotalTime = 34.2563s; SamplesPerSecond = 747.3
-Finished Epoch[16 of 160]: [Training Set] TrainLossPerSample = 0.38623425; EvalErrPerSample = 0.13339999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9354
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.16'
-Starting Epoch 17: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[17 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.36578236; EvalErr[0]PerSample = 0.12671875; TotalTime = 34.2493s; SamplesPerSecond = 747.5
-Finished Epoch[17 of 160]: [Training Set] TrainLossPerSample = 0.37244275; EvalErrPerSample = 0.12964; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9117
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.17'
-Starting Epoch 18: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[18 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.35585205; EvalErr[0]PerSample = 0.12257813; TotalTime = 34.2442s; SamplesPerSecond = 747.6
-Finished Epoch[18 of 160]: [Training Set] TrainLossPerSample = 0.3557232; EvalErrPerSample = 0.12323999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9115
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.18'
-Starting Epoch 19: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[19 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.34668755; EvalErr[0]PerSample = 0.11957031; TotalTime = 34.2499s; SamplesPerSecond = 747.4
-Finished Epoch[19 of 160]: [Training Set] TrainLossPerSample = 0.34480327; EvalErrPerSample = 0.11832; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9097
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.19'
-Starting Epoch 20: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[20 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.32466316; EvalErr[0]PerSample = 0.11343750; TotalTime = 34.2563s; SamplesPerSecond = 747.3
-Finished Epoch[20 of 160]: [Training Set] TrainLossPerSample = 0.33182383; EvalErrPerSample = 0.11554; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9279
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.20'
-Starting Epoch 21: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[21 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.32254467; EvalErr[0]PerSample = 0.11250000; TotalTime = 34.3100s; SamplesPerSecond = 746.1
-Finished Epoch[21 of 160]: [Training Set] TrainLossPerSample = 0.32774779; EvalErrPerSample = 0.11279999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.964
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.21'
-Starting Epoch 22: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[22 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.31498192; EvalErr[0]PerSample = 0.10781250; TotalTime = 34.2773s; SamplesPerSecond = 746.8
-Finished Epoch[22 of 160]: [Training Set] TrainLossPerSample = 0.31775412; EvalErrPerSample = 0.11024; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9334
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.22'
-Starting Epoch 23: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[23 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30091822; EvalErr[0]PerSample = 0.10332031; TotalTime = 34.2440s; SamplesPerSecond = 747.6
-Finished Epoch[23 of 160]: [Training Set] TrainLossPerSample = 0.30627075; EvalErrPerSample = 0.10588; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8985
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.23'
-Starting Epoch 24: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[24 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.29784729; EvalErr[0]PerSample = 0.10226563; TotalTime = 34.2371s; SamplesPerSecond = 747.7
-Finished Epoch[24 of 160]: [Training Set] TrainLossPerSample = 0.30072364; EvalErrPerSample = 0.10337999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8912
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.24'
-Starting Epoch 25: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[25 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28111343; EvalErr[0]PerSample = 0.09546875; TotalTime = 34.2373s; SamplesPerSecond = 747.7
-Finished Epoch[25 of 160]: [Training Set] TrainLossPerSample = 0.29015899; EvalErrPerSample = 0.09956; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8986
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.25'
-Starting Epoch 26: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[26 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28119558; EvalErr[0]PerSample = 0.09894531; TotalTime = 34.2398s; SamplesPerSecond = 747.7
-Finished Epoch[26 of 160]: [Training Set] TrainLossPerSample = 0.28062734; EvalErrPerSample = 0.098979995; AvgLearningRatePerSample = 0.0078125; EpochTime=66.896
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.26'
-Starting Epoch 27: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[27 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.26773546; EvalErr[0]PerSample = 0.09253906; TotalTime = 34.2453s; SamplesPerSecond = 747.5
-Finished Epoch[27 of 160]: [Training Set] TrainLossPerSample = 0.27523825; EvalErrPerSample = 0.095259994; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8905
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.27'
-Starting Epoch 28: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[28 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27260826; EvalErr[0]PerSample = 0.09472656; TotalTime = 34.2339s; SamplesPerSecond = 747.8
-Finished Epoch[28 of 160]: [Training Set] TrainLossPerSample = 0.27595448; EvalErrPerSample = 0.095739998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8925
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.28'
-Starting Epoch 29: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[29 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25796089; EvalErr[0]PerSample = 0.08808594; TotalTime = 34.2304s; SamplesPerSecond = 747.9
-Finished Epoch[29 of 160]: [Training Set] TrainLossPerSample = 0.26468921; EvalErrPerSample = 0.090640001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8905
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.29'
-Starting Epoch 30: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[30 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25212626; EvalErr[0]PerSample = 0.08828125; TotalTime = 34.2402s; SamplesPerSecond = 747.7
-Finished Epoch[30 of 160]: [Training Set] TrainLossPerSample = 0.25851542; EvalErrPerSample = 0.090559997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.889
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.30'
-Starting Epoch 31: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[31 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24511749; EvalErr[0]PerSample = 0.08714844; TotalTime = 34.2452s; SamplesPerSecond = 747.6
-Finished Epoch[31 of 160]: [Training Set] TrainLossPerSample = 0.25035423; EvalErrPerSample = 0.088579997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9067
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.31'
-Starting Epoch 32: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[32 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24218447; EvalErr[0]PerSample = 0.08449219; TotalTime = 34.2393s; SamplesPerSecond = 747.7
-Finished Epoch[32 of 160]: [Training Set] TrainLossPerSample = 0.24899355; EvalErrPerSample = 0.085919999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8996
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.32'
-Starting Epoch 33: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[33 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23380465; EvalErr[0]PerSample = 0.08230469; TotalTime = 34.2486s; SamplesPerSecond = 747.5
-Finished Epoch[33 of 160]: [Training Set] TrainLossPerSample = 0.24616794; EvalErrPerSample = 0.086580001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9199
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.33'
-Starting Epoch 34: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[34 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24045441; EvalErr[0]PerSample = 0.08265625; TotalTime = 34.2421s; SamplesPerSecond = 747.6
-Finished Epoch[34 of 160]: [Training Set] TrainLossPerSample = 0.24212448; EvalErrPerSample = 0.083119996; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9056
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.34'
-Starting Epoch 35: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[35 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.22943111; EvalErr[0]PerSample = 0.08085937; TotalTime = 34.3397s; SamplesPerSecond = 745.5
-Finished Epoch[35 of 160]: [Training Set] TrainLossPerSample = 0.23983407; EvalErrPerSample = 0.083459996; AvgLearningRatePerSample = 0.0078125; EpochTime=67.0061
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.35'
-Starting Epoch 36: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[36 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23311565; EvalErr[0]PerSample = 0.08097656; TotalTime = 34.2423s; SamplesPerSecond = 747.6
-Finished Epoch[36 of 160]: [Training Set] TrainLossPerSample = 0.23720597; EvalErrPerSample = 0.081979997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.905
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.36'
-Starting Epoch 37: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[37 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.21555502; EvalErr[0]PerSample = 0.07417969; TotalTime = 34.2320s; SamplesPerSecond = 747.8
-Finished Epoch[37 of 160]: [Training Set] TrainLossPerSample = 0.2297499; EvalErrPerSample = 0.07948; AvgLearningRatePerSample = 0.0078125; EpochTime=66.89
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.37'
-Starting Epoch 38: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[38 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.22408737; EvalErr[0]PerSample = 0.07808594; TotalTime = 34.2398s; SamplesPerSecond = 747.7
-Finished Epoch[38 of 160]: [Training Set] TrainLossPerSample = 0.22645262; EvalErrPerSample = 0.078979999; AvgLearningRatePerSample = 0.0078125; EpochTime=67.0321
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.38'
-Starting Epoch 39: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[39 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.21047318; EvalErr[0]PerSample = 0.07042969; TotalTime = 34.2497s; SamplesPerSecond = 747.5
-Finished Epoch[39 of 160]: [Training Set] TrainLossPerSample = 0.2245833; EvalErrPerSample = 0.076739997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9139
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.39'
-Starting Epoch 40: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[40 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.21677456; EvalErr[0]PerSample = 0.07500000; TotalTime = 34.2403s; SamplesPerSecond = 747.7
-Finished Epoch[40 of 160]: [Training Set] TrainLossPerSample = 0.22170429; EvalErrPerSample = 0.076480001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8962
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.40'
-Starting Epoch 41: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[41 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.21649546; EvalErr[0]PerSample = 0.07519531; TotalTime = 34.2476s; SamplesPerSecond = 747.5
-Finished Epoch[41 of 160]: [Training Set] TrainLossPerSample = 0.21857023; EvalErrPerSample = 0.075599998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9011
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.41'
-Starting Epoch 42: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[42 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.20455832; EvalErr[0]PerSample = 0.07152344; TotalTime = 34.2415s; SamplesPerSecond = 747.6
-Finished Epoch[42 of 160]: [Training Set] TrainLossPerSample = 0.2160496; EvalErrPerSample = 0.075199999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9069
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.42'
-Starting Epoch 43: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[43 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.20672878; EvalErr[0]PerSample = 0.06980469; TotalTime = 34.2485s; SamplesPerSecond = 747.5
-Finished Epoch[43 of 160]: [Training Set] TrainLossPerSample = 0.21430331; EvalErrPerSample = 0.07418; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9167
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.43'
-Starting Epoch 44: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[44 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.20699423; EvalErr[0]PerSample = 0.07023438; TotalTime = 34.2524s; SamplesPerSecond = 747.4
-Finished Epoch[44 of 160]: [Training Set] TrainLossPerSample = 0.21219908; EvalErrPerSample = 0.072559997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9169
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.44'
-Starting Epoch 45: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[45 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.19827534; EvalErr[0]PerSample = 0.06875000; TotalTime = 34.2522s; SamplesPerSecond = 747.4
-Finished Epoch[45 of 160]: [Training Set] TrainLossPerSample = 0.20897356; EvalErrPerSample = 0.07192; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9303
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.45'
-Starting Epoch 46: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[46 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.20024132; EvalErr[0]PerSample = 0.06968750; TotalTime = 34.2502s; SamplesPerSecond = 747.4
-Finished Epoch[46 of 160]: [Training Set] TrainLossPerSample = 0.20776483; EvalErrPerSample = 0.072839998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9176
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.46'
-Starting Epoch 47: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[47 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.19332899; EvalErr[0]PerSample = 0.06835938; TotalTime = 34.2464s; SamplesPerSecond = 747.5
-Finished Epoch[47 of 160]: [Training Set] TrainLossPerSample = 0.20251136; EvalErrPerSample = 0.071400002; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9216
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.47'
-Starting Epoch 48: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[48 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.20135883; EvalErr[0]PerSample = 0.07085938; TotalTime = 34.2391s; SamplesPerSecond = 747.7
-Finished Epoch[48 of 160]: [Training Set] TrainLossPerSample = 0.20199671; EvalErrPerSample = 0.070560001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9086
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.48'
-Starting Epoch 49: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[49 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18952045; EvalErr[0]PerSample = 0.06582031; TotalTime = 34.2645s; SamplesPerSecond = 747.1
-Finished Epoch[49 of 160]: [Training Set] TrainLossPerSample = 0.19823113; EvalErrPerSample = 0.068580002; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9231
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.49'
-Starting Epoch 50: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[50 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.19133921; EvalErr[0]PerSample = 0.06726562; TotalTime = 34.2493s; SamplesPerSecond = 747.5
-Finished Epoch[50 of 160]: [Training Set] TrainLossPerSample = 0.2006802; EvalErrPerSample = 0.070699997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9319
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.50'
-Starting Epoch 51: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[51 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.19296997; EvalErr[0]PerSample = 0.06621094; TotalTime = 34.2297s; SamplesPerSecond = 747.9
-Finished Epoch[51 of 160]: [Training Set] TrainLossPerSample = 0.1992881; EvalErrPerSample = 0.069499999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8847
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.51'
-Starting Epoch 52: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[52 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18573498; EvalErr[0]PerSample = 0.06414063; TotalTime = 34.2456s; SamplesPerSecond = 747.5
-Finished Epoch[52 of 160]: [Training Set] TrainLossPerSample = 0.19471373; EvalErrPerSample = 0.067099996; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9424
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.52'
-Starting Epoch 53: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[53 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18396889; EvalErr[0]PerSample = 0.06542969; TotalTime = 34.2888s; SamplesPerSecond = 746.6
-Finished Epoch[53 of 160]: [Training Set] TrainLossPerSample = 0.1925908; EvalErrPerSample = 0.067219995; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9471
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.53'
-Starting Epoch 54: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[54 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18691126; EvalErr[0]PerSample = 0.06562500; TotalTime = 34.2456s; SamplesPerSecond = 747.5
-Finished Epoch[54 of 160]: [Training Set] TrainLossPerSample = 0.18951587; EvalErrPerSample = 0.065919995; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9116
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.54'
-Starting Epoch 55: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[55 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18601879; EvalErr[0]PerSample = 0.06371094; TotalTime = 34.2466s; SamplesPerSecond = 747.5
-Finished Epoch[55 of 160]: [Training Set] TrainLossPerSample = 0.18988976; EvalErrPerSample = 0.066100001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9152
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.55'
-Starting Epoch 56: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[56 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18315941; EvalErr[0]PerSample = 0.06406250; TotalTime = 34.2413s; SamplesPerSecond = 747.6
-Finished Epoch[56 of 160]: [Training Set] TrainLossPerSample = 0.18686067; EvalErrPerSample = 0.065739997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9069
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.56'
-Starting Epoch 57: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[57 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18106951; EvalErr[0]PerSample = 0.06246094; TotalTime = 34.2510s; SamplesPerSecond = 747.4
-Finished Epoch[57 of 160]: [Training Set] TrainLossPerSample = 0.19042736; EvalErrPerSample = 0.06622; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9223
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.57'
-Starting Epoch 58: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[58 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18492083; EvalErr[0]PerSample = 0.06527344; TotalTime = 34.2528s; SamplesPerSecond = 747.4
-Finished Epoch[58 of 160]: [Training Set] TrainLossPerSample = 0.18727326; EvalErrPerSample = 0.064939998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9267
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.58'
-Starting Epoch 59: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[59 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17038435; EvalErr[0]PerSample = 0.05917969; TotalTime = 34.2496s; SamplesPerSecond = 747.5
-Finished Epoch[59 of 160]: [Training Set] TrainLossPerSample = 0.18080178; EvalErrPerSample = 0.062299997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9191
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.59'
-Starting Epoch 60: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[60 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17502851; EvalErr[0]PerSample = 0.06183594; TotalTime = 34.2793s; SamplesPerSecond = 746.8
-Finished Epoch[60 of 160]: [Training Set] TrainLossPerSample = 0.17897207; EvalErrPerSample = 0.062819995; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9611
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.60'
-Starting Epoch 61: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[61 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17479221; EvalErr[0]PerSample = 0.06187500; TotalTime = 34.2481s; SamplesPerSecond = 747.5
-Finished Epoch[61 of 160]: [Training Set] TrainLossPerSample = 0.17798932; EvalErrPerSample = 0.062660001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9219
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.61'
-Starting Epoch 62: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[62 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17981518; EvalErr[0]PerSample = 0.06199219; TotalTime = 34.2561s; SamplesPerSecond = 747.3
-Finished Epoch[62 of 160]: [Training Set] TrainLossPerSample = 0.18243115; EvalErrPerSample = 0.064240001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9244
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.62'
-Starting Epoch 63: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[63 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17270412; EvalErr[0]PerSample = 0.05917969; TotalTime = 34.2554s; SamplesPerSecond = 747.3
-Finished Epoch[63 of 160]: [Training Set] TrainLossPerSample = 0.17758152; EvalErrPerSample = 0.061719999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9361
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.63'
-Starting Epoch 64: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[64 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16516579; EvalErr[0]PerSample = 0.05792969; TotalTime = 34.2766s; SamplesPerSecond = 746.9
-Finished Epoch[64 of 160]: [Training Set] TrainLossPerSample = 0.17424898; EvalErrPerSample = 0.06044; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9633
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.64'
-Starting Epoch 65: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[65 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17415672; EvalErr[0]PerSample = 0.06164062; TotalTime = 34.3307s; SamplesPerSecond = 745.7
-Finished Epoch[65 of 160]: [Training Set] TrainLossPerSample = 0.17599966; EvalErrPerSample = 0.062039997; AvgLearningRatePerSample = 0.0078125; EpochTime=67.026
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.65'
-Starting Epoch 66: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[66 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16262278; EvalErr[0]PerSample = 0.05886719; TotalTime = 34.2448s; SamplesPerSecond = 747.6
-Finished Epoch[66 of 160]: [Training Set] TrainLossPerSample = 0.16772267; EvalErrPerSample = 0.059799999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9065
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.66'
-Starting Epoch 67: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[67 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16856216; EvalErr[0]PerSample = 0.05906250; TotalTime = 34.2467s; SamplesPerSecond = 747.5
-Finished Epoch[67 of 160]: [Training Set] TrainLossPerSample = 0.17300676; EvalErrPerSample = 0.061139997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9215
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.67'
-Starting Epoch 68: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[68 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16585386; EvalErr[0]PerSample = 0.05734375; TotalTime = 34.2464s; SamplesPerSecond = 747.5
-Finished Epoch[68 of 160]: [Training Set] TrainLossPerSample = 0.17578007; EvalErrPerSample = 0.060519997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9115
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.68'
-Starting Epoch 69: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[69 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15487971; EvalErr[0]PerSample = 0.05441406; TotalTime = 34.2490s; SamplesPerSecond = 747.5
-Finished Epoch[69 of 160]: [Training Set] TrainLossPerSample = 0.16411984; EvalErrPerSample = 0.05844; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9135
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.69'
-Starting Epoch 70: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[70 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15948469; EvalErr[0]PerSample = 0.05597656; TotalTime = 34.2446s; SamplesPerSecond = 747.6
-Finished Epoch[70 of 160]: [Training Set] TrainLossPerSample = 0.16847; EvalErrPerSample = 0.058879998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9103
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.70'
-Starting Epoch 71: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[71 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16197550; EvalErr[0]PerSample = 0.05718750; TotalTime = 34.2458s; SamplesPerSecond = 747.5
-Finished Epoch[71 of 160]: [Training Set] TrainLossPerSample = 0.17195615; EvalErrPerSample = 0.060839999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9166
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.71'
-Starting Epoch 72: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[72 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15629568; EvalErr[0]PerSample = 0.05515625; TotalTime = 34.2503s; SamplesPerSecond = 747.4
-Finished Epoch[72 of 160]: [Training Set] TrainLossPerSample = 0.17160599; EvalErrPerSample = 0.061219998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9158
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.72'
-Starting Epoch 73: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[73 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15722611; EvalErr[0]PerSample = 0.05476562; TotalTime = 34.2420s; SamplesPerSecond = 747.6
-Finished Epoch[73 of 160]: [Training Set] TrainLossPerSample = 0.16512014; EvalErrPerSample = 0.058259998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8986
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.73'
-Starting Epoch 74: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[74 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16009672; EvalErr[0]PerSample = 0.05703125; TotalTime = 34.2495s; SamplesPerSecond = 747.5
-Finished Epoch[74 of 160]: [Training Set] TrainLossPerSample = 0.16898924; EvalErrPerSample = 0.059019998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9233
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.74'
-Starting Epoch 75: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[75 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15460204; EvalErr[0]PerSample = 0.05312500; TotalTime = 34.2539s; SamplesPerSecond = 747.4
-Finished Epoch[75 of 160]: [Training Set] TrainLossPerSample = 0.16324201; EvalErrPerSample = 0.055919997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9233
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.75'
-Starting Epoch 76: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[76 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15667317; EvalErr[0]PerSample = 0.05515625; TotalTime = 34.2399s; SamplesPerSecond = 747.7
-Finished Epoch[76 of 160]: [Training Set] TrainLossPerSample = 0.15858564; EvalErrPerSample = 0.055059999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9518
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.76'
-Starting Epoch 77: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[77 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15659092; EvalErr[0]PerSample = 0.05410156; TotalTime = 34.2332s; SamplesPerSecond = 747.8
-Finished Epoch[77 of 160]: [Training Set] TrainLossPerSample = 0.16720241; EvalErrPerSample = 0.058079999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9014
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.77'
-Starting Epoch 78: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[78 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15273868; EvalErr[0]PerSample = 0.05164063; TotalTime = 34.2404s; SamplesPerSecond = 747.7
-Finished Epoch[78 of 160]: [Training Set] TrainLossPerSample = 0.15992303; EvalErrPerSample = 0.055239998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.906
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.78'
-Starting Epoch 79: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[79 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15322983; EvalErr[0]PerSample = 0.05476562; TotalTime = 34.2420s; SamplesPerSecond = 747.6
-Finished Epoch[79 of 160]: [Training Set] TrainLossPerSample = 0.16380094; EvalErrPerSample = 0.058139998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9021
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.79'
-Starting Epoch 80: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[80 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15410002; EvalErr[0]PerSample = 0.05335938; TotalTime = 34.2510s; SamplesPerSecond = 747.4
-Finished Epoch[80 of 160]: [Training Set] TrainLossPerSample = 0.16132115; EvalErrPerSample = 0.05638; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9122
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.80'
-Starting Epoch 81: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[81 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.14658320; EvalErr[0]PerSample = 0.05156250; TotalTime = 34.2506s; SamplesPerSecond = 747.4
-Finished Epoch[81 of 160]: [Training Set] TrainLossPerSample = 0.15670048; EvalErrPerSample = 0.055119999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9019
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.81'
-Starting Epoch 82: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[82 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.11055239; EvalErr[0]PerSample = 0.03714844; TotalTime = 34.2453s; SamplesPerSecond = 747.5
-Finished Epoch[82 of 160]: [Training Set] TrainLossPerSample = 0.090582155; EvalErrPerSample = 0.03066; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.8937
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.82'
-Starting Epoch 83: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[83 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05391692; EvalErr[0]PerSample = 0.01675781; TotalTime = 34.2464s; SamplesPerSecond = 747.5
-Finished Epoch[83 of 160]: [Training Set] TrainLossPerSample = 0.052728198; EvalErrPerSample = 0.01644; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9025
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.83'
-Starting Epoch 84: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[84 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04134419; EvalErr[0]PerSample = 0.01242188; TotalTime = 34.2487s; SamplesPerSecond = 747.5
-Finished Epoch[84 of 160]: [Training Set] TrainLossPerSample = 0.04261167; EvalErrPerSample = 0.01278; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.922
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.84'
-Starting Epoch 85: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[85 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03696221; EvalErr[0]PerSample = 0.01214844; TotalTime = 34.2502s; SamplesPerSecond = 747.4
-Finished Epoch[85 of 160]: [Training Set] TrainLossPerSample = 0.035290603; EvalErrPerSample = 0.01108; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9713
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.85'
-Starting Epoch 86: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[86 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03103620; EvalErr[0]PerSample = 0.00906250; TotalTime = 34.2539s; SamplesPerSecond = 747.4
-Finished Epoch[86 of 160]: [Training Set] TrainLossPerSample = 0.031053338; EvalErrPerSample = 0.0093200002; AvgLearningRatePerSample = 0.00078125001; EpochTime=67.0851
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.86'
-Starting Epoch 87: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[87 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02747848; EvalErr[0]PerSample = 0.00800781; TotalTime = 34.9682s; SamplesPerSecond = 732.1
-Finished Epoch[87 of 160]: [Training Set] TrainLossPerSample = 0.028250355; EvalErrPerSample = 0.00832; AvgLearningRatePerSample = 0.00078125001; EpochTime=67.6314
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.87'
-Starting Epoch 88: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[88 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02383066; EvalErr[0]PerSample = 0.00714844; TotalTime = 34.2475s; SamplesPerSecond = 747.5
-Finished Epoch[88 of 160]: [Training Set] TrainLossPerSample = 0.023733234; EvalErrPerSample = 0.0070599997; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9154
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.88'
-Starting Epoch 89: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[89 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02334737; EvalErr[0]PerSample = 0.00632813; TotalTime = 34.2406s; SamplesPerSecond = 747.7
-Finished Epoch[89 of 160]: [Training Set] TrainLossPerSample = 0.022844482; EvalErrPerSample = 0.0063999998; AvgLearningRatePerSample = 0.00078125001; EpochTime=67.0741
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.89'
-Starting Epoch 90: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[90 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02094710; EvalErr[0]PerSample = 0.00589844; TotalTime = 34.2465s; SamplesPerSecond = 747.5
-Finished Epoch[90 of 160]: [Training Set] TrainLossPerSample = 0.020934064; EvalErrPerSample = 0.0059599997; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9229
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.90'
-Starting Epoch 91: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[91 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02006884; EvalErr[0]PerSample = 0.00628906; TotalTime = 34.2548s; SamplesPerSecond = 747.3
-Finished Epoch[91 of 160]: [Training Set] TrainLossPerSample = 0.019843206; EvalErrPerSample = 0.0061399997; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.914
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.91'
-Starting Epoch 92: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[92 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01826200; EvalErr[0]PerSample = 0.00542969; TotalTime = 34.2810s; SamplesPerSecond = 746.8
-Finished Epoch[92 of 160]: [Training Set] TrainLossPerSample = 0.018547757; EvalErrPerSample = 0.00538; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9512
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.92'
-Starting Epoch 93: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[93 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01654859; EvalErr[0]PerSample = 0.00453125; TotalTime = 34.2943s; SamplesPerSecond = 746.5
-Finished Epoch[93 of 160]: [Training Set] TrainLossPerSample = 0.01722255; EvalErrPerSample = 0.0049399999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9701
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.93'
-Starting Epoch 94: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[94 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01585697; EvalErr[0]PerSample = 0.00441406; TotalTime = 34.2504s; SamplesPerSecond = 747.4
-Finished Epoch[94 of 160]: [Training Set] TrainLossPerSample = 0.01515844; EvalErrPerSample = 0.0041799997; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.93
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.94'
-Starting Epoch 95: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[95 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01532144; EvalErr[0]PerSample = 0.00441406; TotalTime = 34.2512s; SamplesPerSecond = 747.4
-Finished Epoch[95 of 160]: [Training Set] TrainLossPerSample = 0.015015967; EvalErrPerSample = 0.0044; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9121
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.95'
-Starting Epoch 96: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[96 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01399448; EvalErr[0]PerSample = 0.00406250; TotalTime = 34.2433s; SamplesPerSecond = 747.6
-Finished Epoch[96 of 160]: [Training Set] TrainLossPerSample = 0.013210027; EvalErrPerSample = 0.00376; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.8997
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.96'
-Starting Epoch 97: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[97 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01209825; EvalErr[0]PerSample = 0.00332031; TotalTime = 34.2567s; SamplesPerSecond = 747.3
-Finished Epoch[97 of 160]: [Training Set] TrainLossPerSample = 0.012226926; EvalErrPerSample = 0.0032599999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9247
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.97'
-Starting Epoch 98: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[98 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01230964; EvalErr[0]PerSample = 0.00343750; TotalTime = 34.2486s; SamplesPerSecond = 747.5
-Finished Epoch[98 of 160]: [Training Set] TrainLossPerSample = 0.012122833; EvalErrPerSample = 0.0034599998; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9192
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.98'
-Starting Epoch 99: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[99 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00988823; EvalErr[0]PerSample = 0.00230469; TotalTime = 34.2477s; SamplesPerSecond = 747.5
-Finished Epoch[99 of 160]: [Training Set] TrainLossPerSample = 0.011153033; EvalErrPerSample = 0.00288; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9079
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.99'
-Starting Epoch 100: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[100 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01078802; EvalErr[0]PerSample = 0.00328125; TotalTime = 34.2446s; SamplesPerSecond = 747.6
-Finished Epoch[100 of 160]: [Training Set] TrainLossPerSample = 0.010805478; EvalErrPerSample = 0.0032199998; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9141
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.100'
-Starting Epoch 101: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[101 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01043408; EvalErr[0]PerSample = 0.00285156; TotalTime = 34.2448s; SamplesPerSecond = 747.6
-Finished Epoch[101 of 160]: [Training Set] TrainLossPerSample = 0.0097289206; EvalErrPerSample = 0.00254; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9115
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.101'
-Starting Epoch 102: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[102 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01028690; EvalErr[0]PerSample = 0.00269531; TotalTime = 34.2500s; SamplesPerSecond = 747.4
-Finished Epoch[102 of 160]: [Training Set] TrainLossPerSample = 0.010017196; EvalErrPerSample = 0.0026999998; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9075
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.102'
-Starting Epoch 103: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[103 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01031995; EvalErr[0]PerSample = 0.00316406; TotalTime = 34.2485s; SamplesPerSecond = 747.5
-Finished Epoch[103 of 160]: [Training Set] TrainLossPerSample = 0.010233736; EvalErrPerSample = 0.0030199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9094
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.103'
-Starting Epoch 104: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[104 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00783067; EvalErr[0]PerSample = 0.00164062; TotalTime = 34.2752s; SamplesPerSecond = 746.9
-Finished Epoch[104 of 160]: [Training Set] TrainLossPerSample = 0.0086551448; EvalErrPerSample = 0.00208; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9965
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.104'
-Starting Epoch 105: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[105 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00939923; EvalErr[0]PerSample = 0.00253906; TotalTime = 34.2407s; SamplesPerSecond = 747.6
-Finished Epoch[105 of 160]: [Training Set] TrainLossPerSample = 0.0096224733; EvalErrPerSample = 0.0026199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.8974
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.105'
-Starting Epoch 106: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[106 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00753223; EvalErr[0]PerSample = 0.00179688; TotalTime = 34.2848s; SamplesPerSecond = 746.7
-Finished Epoch[106 of 160]: [Training Set] TrainLossPerSample = 0.0078934198; EvalErrPerSample = 0.00202; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9655
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.106'
-Starting Epoch 107: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[107 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00865439; EvalErr[0]PerSample = 0.00238281; TotalTime = 34.2544s; SamplesPerSecond = 747.4
-Finished Epoch[107 of 160]: [Training Set] TrainLossPerSample = 0.008151643; EvalErrPerSample = 0.00214; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9284
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.107'
-Starting Epoch 108: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[108 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00694811; EvalErr[0]PerSample = 0.00148438; TotalTime = 34.2490s; SamplesPerSecond = 747.5
-Finished Epoch[108 of 160]: [Training Set] TrainLossPerSample = 0.0073725204; EvalErrPerSample = 0.0016; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9107
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.108'
-Starting Epoch 109: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[109 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00807542; EvalErr[0]PerSample = 0.00218750; TotalTime = 34.2420s; SamplesPerSecond = 747.6
-Finished Epoch[109 of 160]: [Training Set] TrainLossPerSample = 0.0075115636; EvalErrPerSample = 0.00194; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9172
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.109'
-Starting Epoch 110: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[110 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00747269; EvalErr[0]PerSample = 0.00214844; TotalTime = 34.2519s; SamplesPerSecond = 747.4
-Finished Epoch[110 of 160]: [Training Set] TrainLossPerSample = 0.0070967474; EvalErrPerSample = 0.0017599999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9183
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.110'
-Starting Epoch 111: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[111 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00696611; EvalErr[0]PerSample = 0.00175781; TotalTime = 34.2387s; SamplesPerSecond = 747.7
-Finished Epoch[111 of 160]: [Training Set] TrainLossPerSample = 0.0071680127; EvalErrPerSample = 0.0018399999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9065
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.111'
-Starting Epoch 112: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[112 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00660637; EvalErr[0]PerSample = 0.00167969; TotalTime = 34.2517s; SamplesPerSecond = 747.4
-Finished Epoch[112 of 160]: [Training Set] TrainLossPerSample = 0.0069257859; EvalErrPerSample = 0.0018399999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9221
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.112'
-Starting Epoch 113: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[113 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00683305; EvalErr[0]PerSample = 0.00179688; TotalTime = 34.2447s; SamplesPerSecond = 747.6
-Finished Epoch[113 of 160]: [Training Set] TrainLossPerSample = 0.0071453024; EvalErrPerSample = 0.00182; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9078
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.113'
-Starting Epoch 114: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[114 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00659715; EvalErr[0]PerSample = 0.00164062; TotalTime = 34.2397s; SamplesPerSecond = 747.7
-Finished Epoch[114 of 160]: [Training Set] TrainLossPerSample = 0.0066223624; EvalErrPerSample = 0.0017199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9022
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.114'
-Starting Epoch 115: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[115 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00571319; EvalErr[0]PerSample = 0.00140625; TotalTime = 34.2379s; SamplesPerSecond = 747.7
-Finished Epoch[115 of 160]: [Training Set] TrainLossPerSample = 0.0061274339; EvalErrPerSample = 0.0015199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9117
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.115'
-Starting Epoch 116: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[116 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00616163; EvalErr[0]PerSample = 0.00164062; TotalTime = 34.2441s; SamplesPerSecond = 747.6
-Finished Epoch[116 of 160]: [Training Set] TrainLossPerSample = 0.0064637884; EvalErrPerSample = 0.00166; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.92
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.116'
-Starting Epoch 117: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[117 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00667505; EvalErr[0]PerSample = 0.00171875; TotalTime = 34.2353s; SamplesPerSecond = 747.8
-Finished Epoch[117 of 160]: [Training Set] TrainLossPerSample = 0.0063761352; EvalErrPerSample = 0.0016; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.8921
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.117'
-Starting Epoch 118: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[118 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00584337; EvalErr[0]PerSample = 0.00144531; TotalTime = 34.2344s; SamplesPerSecond = 747.8
-Finished Epoch[118 of 160]: [Training Set] TrainLossPerSample = 0.0061604949; EvalErrPerSample = 0.00162; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9005
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.118'
-Starting Epoch 119: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[119 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00587101; EvalErr[0]PerSample = 0.00132813; TotalTime = 34.2448s; SamplesPerSecond = 747.6
-Finished Epoch[119 of 160]: [Training Set] TrainLossPerSample = 0.0058231312; EvalErrPerSample = 0.0013799999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9065
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.119'
-Starting Epoch 120: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[120 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00556109; EvalErr[0]PerSample = 0.00152344; TotalTime = 34.2181s; SamplesPerSecond = 748.1
-Finished Epoch[120 of 160]: [Training Set] TrainLossPerSample = 0.0049992148; EvalErrPerSample = 0.0013; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9296
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.120'
-Starting Epoch 121: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[121 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00544529; EvalErr[0]PerSample = 0.00148438; TotalTime = 34.2507s; SamplesPerSecond = 747.4
-Finished Epoch[121 of 160]: [Training Set] TrainLossPerSample = 0.0059123267; EvalErrPerSample = 0.0016999999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9099
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.121'
-Starting Epoch 122: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[122 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00506212; EvalErr[0]PerSample = 0.00109375; TotalTime = 34.2380s; SamplesPerSecond = 747.7
-Finished Epoch[122 of 160]: [Training Set] TrainLossPerSample = 0.0047056451; EvalErrPerSample = 0.00099999993; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8882
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.122'
-Starting Epoch 123: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[123 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00436568; EvalErr[0]PerSample = 0.00093750; TotalTime = 34.2202s; SamplesPerSecond = 748.1
-Finished Epoch[123 of 160]: [Training Set] TrainLossPerSample = 0.0045333277; EvalErrPerSample = 0.00102; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8649
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.123'
-Starting Epoch 124: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[124 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00423286; EvalErr[0]PerSample = 0.00089844; TotalTime = 34.2368s; SamplesPerSecond = 747.7
-Finished Epoch[124 of 160]: [Training Set] TrainLossPerSample = 0.0040925727; EvalErrPerSample = 0.00078; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8901
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.124'
-Starting Epoch 125: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[125 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00428819; EvalErr[0]PerSample = 0.00089844; TotalTime = 34.2302s; SamplesPerSecond = 747.9
-Finished Epoch[125 of 160]: [Training Set] TrainLossPerSample = 0.0039790319; EvalErrPerSample = 0.00083999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8868
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.125'
-Starting Epoch 126: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[126 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00386990; EvalErr[0]PerSample = 0.00070312; TotalTime = 34.2288s; SamplesPerSecond = 747.9
-Finished Epoch[126 of 160]: [Training Set] TrainLossPerSample = 0.0041431645; EvalErrPerSample = 0.00102; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8935
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.126'
-Starting Epoch 127: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[127 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00415913; EvalErr[0]PerSample = 0.00074219; TotalTime = 34.2434s; SamplesPerSecond = 747.6
-Finished Epoch[127 of 160]: [Training Set] TrainLossPerSample = 0.0038879183; EvalErrPerSample = 0.00072000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9178
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.127'
-Starting Epoch 128: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[128 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00381589; EvalErr[0]PerSample = 0.00074219; TotalTime = 34.2352s; SamplesPerSecond = 747.8
-Finished Epoch[128 of 160]: [Training Set] TrainLossPerSample = 0.0037014519; EvalErrPerSample = 0.00065999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8974
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.128'
-Starting Epoch 129: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[129 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00390699; EvalErr[0]PerSample = 0.00093750; TotalTime = 34.2443s; SamplesPerSecond = 747.6
-Finished Epoch[129 of 160]: [Training Set] TrainLossPerSample = 0.0036388342; EvalErrPerSample = 0.00073999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9066
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.129'
-Starting Epoch 130: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[130 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00401969; EvalErr[0]PerSample = 0.00113281; TotalTime = 34.2339s; SamplesPerSecond = 747.8
-Finished Epoch[130 of 160]: [Training Set] TrainLossPerSample = 0.0038497401; EvalErrPerSample = 0.00089999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8803
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.130'
-Starting Epoch 131: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[131 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00371109; EvalErr[0]PerSample = 0.00050781; TotalTime = 34.2393s; SamplesPerSecond = 747.7
-Finished Epoch[131 of 160]: [Training Set] TrainLossPerSample = 0.0036305008; EvalErrPerSample = 0.00055999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9068
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.131'
-Starting Epoch 132: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[132 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00361982; EvalErr[0]PerSample = 0.00082031; TotalTime = 34.2404s; SamplesPerSecond = 747.7
-Finished Epoch[132 of 160]: [Training Set] TrainLossPerSample = 0.0037875406; EvalErrPerSample = 0.00091999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9025
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.132'
-Starting Epoch 133: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[133 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00379993; EvalErr[0]PerSample = 0.00085938; TotalTime = 34.2434s; SamplesPerSecond = 747.6
-Finished Epoch[133 of 160]: [Training Set] TrainLossPerSample = 0.0037811422; EvalErrPerSample = 0.00072000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8997
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.133'
-Starting Epoch 134: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[134 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00324624; EvalErr[0]PerSample = 0.00042969; TotalTime = 34.2358s; SamplesPerSecond = 747.8
-Finished Epoch[134 of 160]: [Training Set] TrainLossPerSample = 0.0031919111; EvalErrPerSample = 0.00052; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8704
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.134'
-Starting Epoch 135: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[135 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00297883; EvalErr[0]PerSample = 0.00035156; TotalTime = 34.2219s; SamplesPerSecond = 748.1
-Finished Epoch[135 of 160]: [Training Set] TrainLossPerSample = 0.0031994823; EvalErrPerSample = 0.00059999997; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9955
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.135'
-Starting Epoch 136: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[136 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00355394; EvalErr[0]PerSample = 0.00066406; TotalTime = 34.9371s; SamplesPerSecond = 732.7
-Finished Epoch[136 of 160]: [Training Set] TrainLossPerSample = 0.0036795642; EvalErrPerSample = 0.00078; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=67.6302
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.136'
-Starting Epoch 137: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[137 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00359524; EvalErr[0]PerSample = 0.00066406; TotalTime = 34.2354s; SamplesPerSecond = 747.8
-Finished Epoch[137 of 160]: [Training Set] TrainLossPerSample = 0.0035311766; EvalErrPerSample = 0.00068; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.891
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.137'
-Starting Epoch 138: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[138 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00343087; EvalErr[0]PerSample = 0.00066406; TotalTime = 34.2621s; SamplesPerSecond = 747.2
-Finished Epoch[138 of 160]: [Training Set] TrainLossPerSample = 0.0034796586; EvalErrPerSample = 0.00065999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.916
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.138'
-Starting Epoch 139: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[139 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00343002; EvalErr[0]PerSample = 0.00066406; TotalTime = 34.2436s; SamplesPerSecond = 747.6
-Finished Epoch[139 of 160]: [Training Set] TrainLossPerSample = 0.0033328664; EvalErrPerSample = 0.00055999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.923
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.139'
-Starting Epoch 140: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[140 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00337591; EvalErr[0]PerSample = 0.00070312; TotalTime = 34.2353s; SamplesPerSecond = 747.8
-Finished Epoch[140 of 160]: [Training Set] TrainLossPerSample = 0.0031642318; EvalErrPerSample = 0.00055999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.926
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.140'
-Starting Epoch 141: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[141 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00364618; EvalErr[0]PerSample = 0.00078125; TotalTime = 34.2214s; SamplesPerSecond = 748.1
-Finished Epoch[141 of 160]: [Training Set] TrainLossPerSample = 0.0035639035; EvalErrPerSample = 0.00078; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8739
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.141'
-Starting Epoch 142: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[142 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00340368; EvalErr[0]PerSample = 0.00046875; TotalTime = 34.2602s; SamplesPerSecond = 747.2
-Finished Epoch[142 of 160]: [Training Set] TrainLossPerSample = 0.0033217999; EvalErrPerSample = 0.00043999997; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9173
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.142'
-Starting Epoch 143: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[143 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00338313; EvalErr[0]PerSample = 0.00058594; TotalTime = 34.2287s; SamplesPerSecond = 747.9
-Finished Epoch[143 of 160]: [Training Set] TrainLossPerSample = 0.003285937; EvalErrPerSample = 0.00055999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8929
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.143'
-Starting Epoch 144: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[144 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00350241; EvalErr[0]PerSample = 0.00097656; TotalTime = 34.2369s; SamplesPerSecond = 747.7
-Finished Epoch[144 of 160]: [Training Set] TrainLossPerSample = 0.0033956808; EvalErrPerSample = 0.00087999995; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9109
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.144'
-Starting Epoch 145: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[145 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00389780; EvalErr[0]PerSample = 0.00078125; TotalTime = 34.2499s; SamplesPerSecond = 747.4
-Finished Epoch[145 of 160]: [Training Set] TrainLossPerSample = 0.003525309; EvalErrPerSample = 0.00073999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9108
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.145'
-Starting Epoch 146: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[146 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00273976; EvalErr[0]PerSample = 0.00050781; TotalTime = 34.2481s; SamplesPerSecond = 747.5
-Finished Epoch[146 of 160]: [Training Set] TrainLossPerSample = 0.0029731863; EvalErrPerSample = 0.00049999997; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9057
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.146'
-Starting Epoch 147: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[147 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00301678; EvalErr[0]PerSample = 0.00054688; TotalTime = 34.2455s; SamplesPerSecond = 747.5
-Finished Epoch[147 of 160]: [Training Set] TrainLossPerSample = 0.0028605016; EvalErrPerSample = 0.00055999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9068
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.147'
-Starting Epoch 148: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[148 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00318503; EvalErr[0]PerSample = 0.00078125; TotalTime = 34.2349s; SamplesPerSecond = 747.8
-Finished Epoch[148 of 160]: [Training Set] TrainLossPerSample = 0.0029674848; EvalErrPerSample = 0.00063999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8889
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.148'
-Starting Epoch 149: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[149 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00341538; EvalErr[0]PerSample = 0.00070312; TotalTime = 34.2300s; SamplesPerSecond = 747.9
-Finished Epoch[149 of 160]: [Training Set] TrainLossPerSample = 0.0034138309; EvalErrPerSample = 0.00062000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.973
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.149'
-Starting Epoch 150: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[150 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00312641; EvalErr[0]PerSample = 0.00050781; TotalTime = 34.2394s; SamplesPerSecond = 747.7
-Finished Epoch[150 of 160]: [Training Set] TrainLossPerSample = 0.0032544835; EvalErrPerSample = 0.00062000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8886
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.150'
-Starting Epoch 151: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[151 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00294002; EvalErr[0]PerSample = 0.00046875; TotalTime = 34.2440s; SamplesPerSecond = 747.6
-Finished Epoch[151 of 160]: [Training Set] TrainLossPerSample = 0.0029587755; EvalErrPerSample = 0.00049999997; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9165
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.151'
-Starting Epoch 152: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[152 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00312315; EvalErr[0]PerSample = 0.00070312; TotalTime = 34.2407s; SamplesPerSecond = 747.6
-Finished Epoch[152 of 160]: [Training Set] TrainLossPerSample = 0.0031671789; EvalErrPerSample = 0.00063999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8987
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.152'
-Starting Epoch 153: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[153 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00293371; EvalErr[0]PerSample = 0.00042969; TotalTime = 34.2360s; SamplesPerSecond = 747.8
-Finished Epoch[153 of 160]: [Training Set] TrainLossPerSample = 0.0029186283; EvalErrPerSample = 0.00036000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8934
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.153'
-Starting Epoch 154: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[154 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00270366; EvalErr[0]PerSample = 0.00046875; TotalTime = 34.2303s; SamplesPerSecond = 747.9
-Finished Epoch[154 of 160]: [Training Set] TrainLossPerSample = 0.0027337885; EvalErrPerSample = 0.00029999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8932
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.154'
-Starting Epoch 155: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[155 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00262628; EvalErr[0]PerSample = 0.00027344; TotalTime = 34.2340s; SamplesPerSecond = 747.8
-Finished Epoch[155 of 160]: [Training Set] TrainLossPerSample = 0.002633905; EvalErrPerSample = 0.00023999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8889
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.155'
-Starting Epoch 156: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[156 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00307139; EvalErr[0]PerSample = 0.00074219; TotalTime = 34.2388s; SamplesPerSecond = 747.7
-Finished Epoch[156 of 160]: [Training Set] TrainLossPerSample = 0.0029770101; EvalErrPerSample = 0.00065999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8912
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.156'
-Starting Epoch 157: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[157 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00275711; EvalErr[0]PerSample = 0.00031250; TotalTime = 34.2329s; SamplesPerSecond = 747.8
-Finished Epoch[157 of 160]: [Training Set] TrainLossPerSample = 0.0026996653; EvalErrPerSample = 0.00034; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8768
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.157'
-Starting Epoch 158: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[158 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00278720; EvalErr[0]PerSample = 0.00042969; TotalTime = 34.2421s; SamplesPerSecond = 747.6
-Finished Epoch[158 of 160]: [Training Set] TrainLossPerSample = 0.0026968825; EvalErrPerSample = 0.00037999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9344
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.158'
-Starting Epoch 159: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[159 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00272500; EvalErr[0]PerSample = 0.00031250; TotalTime = 34.2259s; SamplesPerSecond = 748.0
-Finished Epoch[159 of 160]: [Training Set] TrainLossPerSample = 0.0027625638; EvalErrPerSample = 0.00037999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9162
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56.159'
-Starting Epoch 160: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
-
-Starting minibatch loop.
- Epoch[160 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00257069; EvalErr[0]PerSample = 0.00042969; TotalTime = 34.2364s; SamplesPerSecond = 747.7
-Finished Epoch[160 of 160]: [Training Set] TrainLossPerSample = 0.0029211191; EvalErrPerSample = 0.00063999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8957
-SGD: Saving checkpoint model './Output/Models/04_ResNet_56'
-CNTKCommandTrainEnd: Train
-
-Post-processing network...
-
-3 roots:
-	CE = CrossEntropyWithSoftmax
-	Err = ErrorPrediction
-	OutputNodes.z = Plus
-FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
-FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
-FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
-
-
-Validating network. 949 nodes to process in pass 1.
-
-Validating --> labels = InputValue -> [10 x *]
-Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
-Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
-Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
-Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
-Validating --> features = InputValue -> [32 x 32 x 3 x *]
-Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
-Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
-Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
-Validating --> OutputNodes.b = LearnableParameter -> [10]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-
-Validating network. 390 nodes to process in pass 2.
-
-Validating --> labels = InputValue -> [10 x *]
-Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
-Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
-Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
-Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
-Validating --> features = InputValue -> [32 x 32 x 3 x *]
-Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
-Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
-Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
-Validating --> OutputNodes.b = LearnableParameter -> [10]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-
-Validating network, final pass.
-
-Validating --> labels = InputValue -> [10 x *]
-Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
-Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
-Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
-Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
-Validating --> features = InputValue -> [32 x 32 x 3 x *]
-Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
-Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
-Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
-Validating --> OutputNodes.b = LearnableParameter -> [10]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-
-559 out of 949 nodes do not share the minibatch layout with the input data.
-
-Post-processing network complete.
-
-Post-processing network...
-
-3 roots:
-	CE = CrossEntropyWithSoftmax
-	Err = ErrorPrediction
-	OutputNodes.z = Plus
-FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
-FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
-FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
-
-
-Validating network. 949 nodes to process in pass 1.
-
-Validating --> labels = InputValue -> [10 x *]
-Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
-Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
-Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
-Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
-Validating --> features = InputValue -> [32 x 32 x 3 x *]
-Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
-Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
-Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
-Validating --> OutputNodes.b = LearnableParameter -> [10]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-
-Validating network. 390 nodes to process in pass 2.
-
-Validating --> labels = InputValue -> [10 x *]
-Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
-Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
-Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
-Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
-Validating --> features = InputValue -> [32 x 32 x 3 x *]
-Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
-Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
-Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
-Validating --> OutputNodes.b = LearnableParameter -> [10]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-
-Validating network, final pass.
-
-Validating --> labels = InputValue -> [10 x *]
-Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
-Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
-Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
-Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
-Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
-Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
-Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
-Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
-Validating --> features = InputValue -> [32 x 32 x 3 x *]
-Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
-Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
-Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
-Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
-Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
-Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
-Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
-Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
-Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
-Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
-Validating --> OutputNodes.b = LearnableParameter -> [10]
-Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
-Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
-
-559 out of 949 nodes do not share the minibatch layout with the input data.
-
-Post-processing network complete.
-evalNodeNames are not specified, using all the default evalnodes and training criterion nodes.
-
-
-Allocating matrices for forward and/or backward propagation.
-Minibatch[1-20]: Samples Seen = 10000    Err: ErrorPrediction/Sample = 0.0644    CE: CrossEntropyWithSoftmax/Sample = 0.3034767    
-Final Results: Minibatch[1-20]: Samples Seen = 10000    Err: ErrorPrediction/Sample = 0.0644    CE: CrossEntropyWithSoftmax/Sample = 0.3034767    Perplexity = 1.35456    
-COMPLETED
+-------------------------------------------------------------------
+Build info: 
+
+		Built time: Feb 26 2016 16:36:12
+		Last modified date: Thu Feb 25 12:56:12 2016
+		Build type: Release
+		Build target: GPU
+		With 1bit-SGD: no
+		CUDA_PATH: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5
+		CUB_PATH: C:\src\cub
+		CUDNN_PATH: C:\NVIDIA\cudnn-4.0\cuda
+		Build Branch: 
+		Build SHA1:  (modified)
+		Built by alexeyk on z840-01
+		Build Path: C:\src\cntk\Source\CNTK\
+-------------------------------------------------------------------
+running on z840-01 at 2016/02/26 17:43:42
+command line: 
+C:\src\cntk\x64\Release\CNTK.exe configFile=04_ResNet_56.cntk 
+
+>>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>
+RootDir = "."
+ConfigDir = "$RootDir$"
+DataDir = "$RootDir$"
+OutputDir = "$RootDir$/Output"
+ModelDir = "$OutputDir$/Models"
+ndlMacros="$ConfigDir$/Macros.ndl"
+precision="float"
+deviceId="Auto"
+prefetch="true"
+parallelTrain="false"
+command=Train:AddBNEval:Test
+stderr="$OutputDir$/04_ResNet_56"
+traceLevel=1
+numMBsToShowResult=200
+Proj16to32Filename = "$ConfigDir$/16to32.txt"
+Proj32to64Filename = "$ConfigDir$/32to64.txt"
+Train=[
+    action="train"
+    modelPath="$ModelDir$/04_ResNet_56"
+     NDLNetworkBuilder=[
+        networkDescription="$ConfigDir$/04_ResNet_56.ndl"
+    ]
+    SGD=[
+        epochSize=0
+        minibatchSize=128
+        learningRatesPerMB=0.1*1:1.0*80:0.1*40:0.01
+        momentumPerMB=0.9
+        maxEpochs=160
+        L2RegWeight=0.0001
+        dropoutRate=0
+        ParallelTrain=[
+            parallelizationMethod="DataParallelSGD"
+            distributedMBReading="true"
+            parallelizationStartEpoch=1
+            DataParallelSGD=[
+                gradientBits=32
+            ]
+        ]
+    ]
+    reader=[
+        readerType="ImageReader"
+        file="$DataDir$/train_map.txt"
+        randomize="Auto"
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType="Random"
+            cropRatio=0.8
+            jitterType="UniRatio"
+            interpolations="Linear"
+            meanFile="$ConfigDir$/CIFAR-10_mean.xml"
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+AddBNEval=[    
+    action="edit"
+    CurModel="$ModelDir$/04_ResNet_56"
+    NewModel="$ModelDir$/04_ResNet_56.Eval"
+    editPath="$ConfigDir$/03_ResNet.mel"
+]
+Test=[
+    action="test"
+    modelPath="$ModelDir$/04_ResNet_56.Eval"
+    minibatchSize=512
+    reader=[
+        readerType="ImageReader"
+        file="$DataDir$/test_map.txt"
+        randomize="None"
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType="Center"
+            cropRatio=1
+            jitterType="UniRatio"
+            interpolations="Linear"
+            meanFile="$ConfigDir$/CIFAR-10_mean.xml"
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+
+<<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<
+
+>>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
+RootDir = "."
+ConfigDir = "."
+DataDir = "."
+OutputDir = "./Output"
+ModelDir = "./Output/Models"
+ndlMacros="./Macros.ndl"
+precision="float"
+deviceId="Auto"
+prefetch="true"
+parallelTrain="false"
+command=Train:AddBNEval:Test
+stderr="./Output/04_ResNet_56"
+traceLevel=1
+numMBsToShowResult=200
+Proj16to32Filename = "./16to32.txt"
+Proj32to64Filename = "./32to64.txt"
+Train=[
+    action="train"
+    modelPath="./Output/Models/04_ResNet_56"
+     NDLNetworkBuilder=[
+        networkDescription="./04_ResNet_56.ndl"
+    ]
+    SGD=[
+        epochSize=0
+        minibatchSize=128
+        learningRatesPerMB=0.1*1:1.0*80:0.1*40:0.01
+        momentumPerMB=0.9
+        maxEpochs=160
+        L2RegWeight=0.0001
+        dropoutRate=0
+        ParallelTrain=[
+            parallelizationMethod="DataParallelSGD"
+            distributedMBReading="true"
+            parallelizationStartEpoch=1
+            DataParallelSGD=[
+                gradientBits=32
+            ]
+        ]
+    ]
+    reader=[
+        readerType="ImageReader"
+        file="./train_map.txt"
+        randomize="Auto"
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType="Random"
+            cropRatio=0.8
+            jitterType="UniRatio"
+            interpolations="Linear"
+            meanFile="./CIFAR-10_mean.xml"
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+AddBNEval=[    
+    action="edit"
+    CurModel="./Output/Models/04_ResNet_56"
+    NewModel="./Output/Models/04_ResNet_56.Eval"
+    editPath="./03_ResNet.mel"
+]
+Test=[
+    action="test"
+    modelPath="./Output/Models/04_ResNet_56.Eval"
+    minibatchSize=512
+    reader=[
+        readerType="ImageReader"
+        file="./test_map.txt"
+        randomize="None"
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType="Center"
+            cropRatio=1
+            jitterType="UniRatio"
+            interpolations="Linear"
+            meanFile="./CIFAR-10_mean.xml"
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+
+<<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
+
+>>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
+configparameters: 04_ResNet_56.cntk:AddBNEval=[    
+    action="edit"
+    CurModel="./Output/Models/04_ResNet_56"
+    NewModel="./Output/Models/04_ResNet_56.Eval"
+    editPath="./03_ResNet.mel"
+]
+
+configparameters: 04_ResNet_56.cntk:command=Train:AddBNEval:Test
+configparameters: 04_ResNet_56.cntk:ConfigDir=.
+configparameters: 04_ResNet_56.cntk:DataDir=.
+configparameters: 04_ResNet_56.cntk:deviceId=Auto
+configparameters: 04_ResNet_56.cntk:ModelDir=./Output/Models
+configparameters: 04_ResNet_56.cntk:ndlMacros=./Macros.ndl
+configparameters: 04_ResNet_56.cntk:numMBsToShowResult=200
+configparameters: 04_ResNet_56.cntk:OutputDir=./Output
+configparameters: 04_ResNet_56.cntk:parallelTrain=false
+configparameters: 04_ResNet_56.cntk:precision=float
+configparameters: 04_ResNet_56.cntk:prefetch=true
+configparameters: 04_ResNet_56.cntk:Proj16to32Filename=./16to32.txt
+configparameters: 04_ResNet_56.cntk:Proj32to64Filename=./32to64.txt
+configparameters: 04_ResNet_56.cntk:RootDir=.
+configparameters: 04_ResNet_56.cntk:stderr=./Output/04_ResNet_56
+configparameters: 04_ResNet_56.cntk:Test=[
+    action="test"
+    modelPath="./Output/Models/04_ResNet_56.Eval"
+    minibatchSize=512
+    reader=[
+        readerType="ImageReader"
+        file="./test_map.txt"
+        randomize="None"
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType="Center"
+            cropRatio=1
+            jitterType="UniRatio"
+            interpolations="Linear"
+            meanFile="./CIFAR-10_mean.xml"
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+
+configparameters: 04_ResNet_56.cntk:traceLevel=1
+configparameters: 04_ResNet_56.cntk:Train=[
+    action="train"
+    modelPath="./Output/Models/04_ResNet_56"
+     NDLNetworkBuilder=[
+        networkDescription="./04_ResNet_56.ndl"
+    ]
+    SGD=[
+        epochSize=0
+        minibatchSize=128
+        learningRatesPerMB=0.1*1:1.0*80:0.1*40:0.01
+        momentumPerMB=0.9
+        maxEpochs=160
+        L2RegWeight=0.0001
+        dropoutRate=0
+        ParallelTrain=[
+            parallelizationMethod="DataParallelSGD"
+            distributedMBReading="true"
+            parallelizationStartEpoch=1
+            DataParallelSGD=[
+                gradientBits=32
+            ]
+        ]
+    ]
+    reader=[
+        readerType="ImageReader"
+        file="./train_map.txt"
+        randomize="Auto"
+        features=[
+            width=32
+            height=32
+            channels=3
+            cropType="Random"
+            cropRatio=0.8
+            jitterType="UniRatio"
+            interpolations="Linear"
+            meanFile="./CIFAR-10_mean.xml"
+        ]
+        labels=[
+            labelDim=10
+        ]
+    ]    
+]
+
+<<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
+command: Train AddBNEval Test 
+precision = float
+CNTKModelPath: ./Output/Models/04_ResNet_56
+CNTKCommandTrainInfo: Train : 160
+CNTKCommandTrainInfo: CNTKNoMoreCommands_Total : 160
+CNTKCommandTrainBegin: Train
+LockDevice: Locked GPU 0 to test availability.
+LockDevice: Unlocked GPU 0 after testing.
+LockDevice: Locked GPU 1 to test availability.
+LockDevice: Unlocked GPU 1 after testing.
+LockDevice: Locked GPU 2 to test availability.
+LockDevice: Unlocked GPU 2 after testing.
+LockDevice: Locked GPU 0 for exclusive use.
+NDLBuilder Using GPU 0
+Microsoft::MSR::CNTK::GPUMatrix<ElemType>::SetGaussianRandomValue (GPU): creating curand object with seed 1, sizeof(ElemType)==4
+
+Post-processing network...
+
+3 roots:
+	CE = CrossEntropyWithSoftmax
+	Err = ErrorPrediction
+	OutputNodes.z = Plus
+FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
+FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
+FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
+
+
+Validating network. 949 nodes to process in pass 1.
+
+Validating --> labels = InputValue -> [10 x *]
+Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
+Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
+Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
+Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
+Validating --> features = InputValue -> [32 x 32 x 3 x *]
+Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
+Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
+Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
+Validating --> OutputNodes.b = LearnableParameter -> [10]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+
+Validating network. 390 nodes to process in pass 2.
+
+Validating --> labels = InputValue -> [10 x *]
+Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
+Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
+Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
+Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
+Validating --> features = InputValue -> [32 x 32 x 3 x *]
+Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
+Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
+Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
+Validating --> OutputNodes.b = LearnableParameter -> [10]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+
+Validating network, final pass.
+
+Validating --> labels = InputValue -> [10 x *]
+Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
+Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
+Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
+Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
+Validating --> features = InputValue -> [32 x 32 x 3 x *]
+Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
+Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
+Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
+Validating --> OutputNodes.b = LearnableParameter -> [10]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+
+559 out of 949 nodes do not share the minibatch layout with the input data.
+
+Post-processing network complete.
+
+SGD using GPU 0.
+
+Training criterion node(s):
+	CE = CrossEntropyWithSoftmax
+
+Evaluation criterion node(s):
+	Err = ErrorPrediction
+
+
+Allocating matrices for forward and/or backward propagation.
+No PreCompute nodes found, skipping PreCompute step
+Starting Epoch 1: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 1 of 160]-Minibatch[   1- 200]: SamplesSeen = 25600; TrainLossPerSample =  2.08201447; EvalErr[0]PerSample = 0.79109375; TotalTime = 37.3101s; SamplesPerSecond = 686.1
+Finished Epoch[ 1 of 160]: [Training Set] TrainLossPerSample = 1.9284772; EvalErrPerSample = 0.73097998; AvgLearningRatePerSample = 0.00078125001; EpochTime=69.7251
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.1'
+Starting Epoch 2: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 2 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  1.80199310; EvalErr[0]PerSample = 0.68078125; TotalTime = 34.7295s; SamplesPerSecond = 737.1
+Finished Epoch[ 2 of 160]: [Training Set] TrainLossPerSample = 1.6386911; EvalErrPerSample = 0.61193997; AvgLearningRatePerSample = 0.0078125; EpochTime=67.9925
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.2'
+Starting Epoch 3: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 3 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  1.25481010; EvalErr[0]PerSample = 0.45394531; TotalTime = 34.5533s; SamplesPerSecond = 740.9
+Finished Epoch[ 3 of 160]: [Training Set] TrainLossPerSample = 1.1870104; EvalErrPerSample = 0.42773998; AvgLearningRatePerSample = 0.0078125; EpochTime=67.6327
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.3'
+Starting Epoch 4: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 4 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  1.00952705; EvalErr[0]PerSample = 0.36304687; TotalTime = 34.5927s; SamplesPerSecond = 740.0
+Finished Epoch[ 4 of 160]: [Training Set] TrainLossPerSample = 0.97715324; EvalErrPerSample = 0.34739998; AvgLearningRatePerSample = 0.0078125; EpochTime=67.5202
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.4'
+Starting Epoch 5: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 5 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.84818413; EvalErr[0]PerSample = 0.29589844; TotalTime = 34.3363s; SamplesPerSecond = 745.6
+Finished Epoch[ 5 of 160]: [Training Set] TrainLossPerSample = 0.83634055; EvalErrPerSample = 0.29328001; AvgLearningRatePerSample = 0.0078125; EpochTime=67.0961
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.5'
+Starting Epoch 6: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 6 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.74924049; EvalErr[0]PerSample = 0.26066406; TotalTime = 34.2546s; SamplesPerSecond = 747.3
+Finished Epoch[ 6 of 160]: [Training Set] TrainLossPerSample = 0.73574179; EvalErrPerSample = 0.25577998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9549
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.6'
+Starting Epoch 7: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 7 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.66103149; EvalErr[0]PerSample = 0.22968750; TotalTime = 34.2366s; SamplesPerSecond = 747.7
+Finished Epoch[ 7 of 160]: [Training Set] TrainLossPerSample = 0.65649849; EvalErrPerSample = 0.22679999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8964
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.7'
+Starting Epoch 8: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 8 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.59962826; EvalErr[0]PerSample = 0.20585938; TotalTime = 34.2398s; SamplesPerSecond = 747.7
+Finished Epoch[ 8 of 160]: [Training Set] TrainLossPerSample = 0.59828544; EvalErrPerSample = 0.20622; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8987
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.8'
+Starting Epoch 9: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[ 9 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.56160091; EvalErr[0]PerSample = 0.19425781; TotalTime = 34.2457s; SamplesPerSecond = 747.5
+Finished Epoch[ 9 of 160]: [Training Set] TrainLossPerSample = 0.56435287; EvalErrPerSample = 0.19484; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9167
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.9'
+Starting Epoch 10: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[10 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.52267384; EvalErr[0]PerSample = 0.18140625; TotalTime = 34.2539s; SamplesPerSecond = 747.4
+Finished Epoch[10 of 160]: [Training Set] TrainLossPerSample = 0.517959; EvalErrPerSample = 0.17936; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9219
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.10'
+Starting Epoch 11: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[11 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.49345413; EvalErr[0]PerSample = 0.17046875; TotalTime = 34.2380s; SamplesPerSecond = 747.7
+Finished Epoch[11 of 160]: [Training Set] TrainLossPerSample = 0.48473847; EvalErrPerSample = 0.16725999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9081
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.11'
+Starting Epoch 12: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[12 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.46631210; EvalErr[0]PerSample = 0.16121094; TotalTime = 34.2568s; SamplesPerSecond = 747.3
+Finished Epoch[12 of 160]: [Training Set] TrainLossPerSample = 0.46427867; EvalErrPerSample = 0.16023999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.989
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.12'
+Starting Epoch 13: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[13 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.44940582; EvalErr[0]PerSample = 0.15414063; TotalTime = 34.2965s; SamplesPerSecond = 746.4
+Finished Epoch[13 of 160]: [Training Set] TrainLossPerSample = 0.44242746; EvalErrPerSample = 0.15223999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9538
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.13'
+Starting Epoch 14: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[14 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.41690651; EvalErr[0]PerSample = 0.14292969; TotalTime = 34.2325s; SamplesPerSecond = 747.8
+Finished Epoch[14 of 160]: [Training Set] TrainLossPerSample = 0.41924006; EvalErrPerSample = 0.1444; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8936
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.14'
+Starting Epoch 15: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[15 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.39440144; EvalErr[0]PerSample = 0.13656250; TotalTime = 34.2478s; SamplesPerSecond = 747.5
+Finished Epoch[15 of 160]: [Training Set] TrainLossPerSample = 0.39710063; EvalErrPerSample = 0.13676; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9095
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.15'
+Starting Epoch 16: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[16 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.38027695; EvalErr[0]PerSample = 0.13261719; TotalTime = 34.2563s; SamplesPerSecond = 747.3
+Finished Epoch[16 of 160]: [Training Set] TrainLossPerSample = 0.38623425; EvalErrPerSample = 0.13339999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9354
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.16'
+Starting Epoch 17: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[17 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.36578236; EvalErr[0]PerSample = 0.12671875; TotalTime = 34.2493s; SamplesPerSecond = 747.5
+Finished Epoch[17 of 160]: [Training Set] TrainLossPerSample = 0.37244275; EvalErrPerSample = 0.12964; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9117
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.17'
+Starting Epoch 18: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[18 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.35585205; EvalErr[0]PerSample = 0.12257813; TotalTime = 34.2442s; SamplesPerSecond = 747.6
+Finished Epoch[18 of 160]: [Training Set] TrainLossPerSample = 0.3557232; EvalErrPerSample = 0.12323999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9115
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.18'
+Starting Epoch 19: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[19 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.34668755; EvalErr[0]PerSample = 0.11957031; TotalTime = 34.2499s; SamplesPerSecond = 747.4
+Finished Epoch[19 of 160]: [Training Set] TrainLossPerSample = 0.34480327; EvalErrPerSample = 0.11832; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9097
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.19'
+Starting Epoch 20: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[20 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.32466316; EvalErr[0]PerSample = 0.11343750; TotalTime = 34.2563s; SamplesPerSecond = 747.3
+Finished Epoch[20 of 160]: [Training Set] TrainLossPerSample = 0.33182383; EvalErrPerSample = 0.11554; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9279
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.20'
+Starting Epoch 21: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[21 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.32254467; EvalErr[0]PerSample = 0.11250000; TotalTime = 34.3100s; SamplesPerSecond = 746.1
+Finished Epoch[21 of 160]: [Training Set] TrainLossPerSample = 0.32774779; EvalErrPerSample = 0.11279999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.964
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.21'
+Starting Epoch 22: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[22 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.31498192; EvalErr[0]PerSample = 0.10781250; TotalTime = 34.2773s; SamplesPerSecond = 746.8
+Finished Epoch[22 of 160]: [Training Set] TrainLossPerSample = 0.31775412; EvalErrPerSample = 0.11024; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9334
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.22'
+Starting Epoch 23: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[23 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.30091822; EvalErr[0]PerSample = 0.10332031; TotalTime = 34.2440s; SamplesPerSecond = 747.6
+Finished Epoch[23 of 160]: [Training Set] TrainLossPerSample = 0.30627075; EvalErrPerSample = 0.10588; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8985
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.23'
+Starting Epoch 24: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[24 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.29784729; EvalErr[0]PerSample = 0.10226563; TotalTime = 34.2371s; SamplesPerSecond = 747.7
+Finished Epoch[24 of 160]: [Training Set] TrainLossPerSample = 0.30072364; EvalErrPerSample = 0.10337999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8912
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.24'
+Starting Epoch 25: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[25 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28111343; EvalErr[0]PerSample = 0.09546875; TotalTime = 34.2373s; SamplesPerSecond = 747.7
+Finished Epoch[25 of 160]: [Training Set] TrainLossPerSample = 0.29015899; EvalErrPerSample = 0.09956; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8986
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.25'
+Starting Epoch 26: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[26 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.28119558; EvalErr[0]PerSample = 0.09894531; TotalTime = 34.2398s; SamplesPerSecond = 747.7
+Finished Epoch[26 of 160]: [Training Set] TrainLossPerSample = 0.28062734; EvalErrPerSample = 0.098979995; AvgLearningRatePerSample = 0.0078125; EpochTime=66.896
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.26'
+Starting Epoch 27: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[27 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.26773546; EvalErr[0]PerSample = 0.09253906; TotalTime = 34.2453s; SamplesPerSecond = 747.5
+Finished Epoch[27 of 160]: [Training Set] TrainLossPerSample = 0.27523825; EvalErrPerSample = 0.095259994; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8905
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.27'
+Starting Epoch 28: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[28 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.27260826; EvalErr[0]PerSample = 0.09472656; TotalTime = 34.2339s; SamplesPerSecond = 747.8
+Finished Epoch[28 of 160]: [Training Set] TrainLossPerSample = 0.27595448; EvalErrPerSample = 0.095739998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8925
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.28'
+Starting Epoch 29: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[29 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25796089; EvalErr[0]PerSample = 0.08808594; TotalTime = 34.2304s; SamplesPerSecond = 747.9
+Finished Epoch[29 of 160]: [Training Set] TrainLossPerSample = 0.26468921; EvalErrPerSample = 0.090640001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8905
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.29'
+Starting Epoch 30: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[30 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.25212626; EvalErr[0]PerSample = 0.08828125; TotalTime = 34.2402s; SamplesPerSecond = 747.7
+Finished Epoch[30 of 160]: [Training Set] TrainLossPerSample = 0.25851542; EvalErrPerSample = 0.090559997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.889
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.30'
+Starting Epoch 31: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[31 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24511749; EvalErr[0]PerSample = 0.08714844; TotalTime = 34.2452s; SamplesPerSecond = 747.6
+Finished Epoch[31 of 160]: [Training Set] TrainLossPerSample = 0.25035423; EvalErrPerSample = 0.088579997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9067
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.31'
+Starting Epoch 32: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[32 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24218447; EvalErr[0]PerSample = 0.08449219; TotalTime = 34.2393s; SamplesPerSecond = 747.7
+Finished Epoch[32 of 160]: [Training Set] TrainLossPerSample = 0.24899355; EvalErrPerSample = 0.085919999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8996
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.32'
+Starting Epoch 33: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[33 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23380465; EvalErr[0]PerSample = 0.08230469; TotalTime = 34.2486s; SamplesPerSecond = 747.5
+Finished Epoch[33 of 160]: [Training Set] TrainLossPerSample = 0.24616794; EvalErrPerSample = 0.086580001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9199
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.33'
+Starting Epoch 34: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[34 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.24045441; EvalErr[0]PerSample = 0.08265625; TotalTime = 34.2421s; SamplesPerSecond = 747.6
+Finished Epoch[34 of 160]: [Training Set] TrainLossPerSample = 0.24212448; EvalErrPerSample = 0.083119996; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9056
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.34'
+Starting Epoch 35: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[35 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.22943111; EvalErr[0]PerSample = 0.08085937; TotalTime = 34.3397s; SamplesPerSecond = 745.5
+Finished Epoch[35 of 160]: [Training Set] TrainLossPerSample = 0.23983407; EvalErrPerSample = 0.083459996; AvgLearningRatePerSample = 0.0078125; EpochTime=67.0061
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.35'
+Starting Epoch 36: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[36 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.23311565; EvalErr[0]PerSample = 0.08097656; TotalTime = 34.2423s; SamplesPerSecond = 747.6
+Finished Epoch[36 of 160]: [Training Set] TrainLossPerSample = 0.23720597; EvalErrPerSample = 0.081979997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.905
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.36'
+Starting Epoch 37: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[37 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.21555502; EvalErr[0]PerSample = 0.07417969; TotalTime = 34.2320s; SamplesPerSecond = 747.8
+Finished Epoch[37 of 160]: [Training Set] TrainLossPerSample = 0.2297499; EvalErrPerSample = 0.07948; AvgLearningRatePerSample = 0.0078125; EpochTime=66.89
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.37'
+Starting Epoch 38: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[38 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.22408737; EvalErr[0]PerSample = 0.07808594; TotalTime = 34.2398s; SamplesPerSecond = 747.7
+Finished Epoch[38 of 160]: [Training Set] TrainLossPerSample = 0.22645262; EvalErrPerSample = 0.078979999; AvgLearningRatePerSample = 0.0078125; EpochTime=67.0321
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.38'
+Starting Epoch 39: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[39 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.21047318; EvalErr[0]PerSample = 0.07042969; TotalTime = 34.2497s; SamplesPerSecond = 747.5
+Finished Epoch[39 of 160]: [Training Set] TrainLossPerSample = 0.2245833; EvalErrPerSample = 0.076739997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9139
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.39'
+Starting Epoch 40: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[40 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.21677456; EvalErr[0]PerSample = 0.07500000; TotalTime = 34.2403s; SamplesPerSecond = 747.7
+Finished Epoch[40 of 160]: [Training Set] TrainLossPerSample = 0.22170429; EvalErrPerSample = 0.076480001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8962
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.40'
+Starting Epoch 41: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[41 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.21649546; EvalErr[0]PerSample = 0.07519531; TotalTime = 34.2476s; SamplesPerSecond = 747.5
+Finished Epoch[41 of 160]: [Training Set] TrainLossPerSample = 0.21857023; EvalErrPerSample = 0.075599998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9011
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.41'
+Starting Epoch 42: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[42 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.20455832; EvalErr[0]PerSample = 0.07152344; TotalTime = 34.2415s; SamplesPerSecond = 747.6
+Finished Epoch[42 of 160]: [Training Set] TrainLossPerSample = 0.2160496; EvalErrPerSample = 0.075199999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9069
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.42'
+Starting Epoch 43: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[43 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.20672878; EvalErr[0]PerSample = 0.06980469; TotalTime = 34.2485s; SamplesPerSecond = 747.5
+Finished Epoch[43 of 160]: [Training Set] TrainLossPerSample = 0.21430331; EvalErrPerSample = 0.07418; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9167
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.43'
+Starting Epoch 44: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[44 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.20699423; EvalErr[0]PerSample = 0.07023438; TotalTime = 34.2524s; SamplesPerSecond = 747.4
+Finished Epoch[44 of 160]: [Training Set] TrainLossPerSample = 0.21219908; EvalErrPerSample = 0.072559997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9169
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.44'
+Starting Epoch 45: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[45 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.19827534; EvalErr[0]PerSample = 0.06875000; TotalTime = 34.2522s; SamplesPerSecond = 747.4
+Finished Epoch[45 of 160]: [Training Set] TrainLossPerSample = 0.20897356; EvalErrPerSample = 0.07192; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9303
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.45'
+Starting Epoch 46: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[46 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.20024132; EvalErr[0]PerSample = 0.06968750; TotalTime = 34.2502s; SamplesPerSecond = 747.4
+Finished Epoch[46 of 160]: [Training Set] TrainLossPerSample = 0.20776483; EvalErrPerSample = 0.072839998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9176
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.46'
+Starting Epoch 47: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[47 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.19332899; EvalErr[0]PerSample = 0.06835938; TotalTime = 34.2464s; SamplesPerSecond = 747.5
+Finished Epoch[47 of 160]: [Training Set] TrainLossPerSample = 0.20251136; EvalErrPerSample = 0.071400002; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9216
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.47'
+Starting Epoch 48: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[48 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.20135883; EvalErr[0]PerSample = 0.07085938; TotalTime = 34.2391s; SamplesPerSecond = 747.7
+Finished Epoch[48 of 160]: [Training Set] TrainLossPerSample = 0.20199671; EvalErrPerSample = 0.070560001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9086
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.48'
+Starting Epoch 49: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[49 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18952045; EvalErr[0]PerSample = 0.06582031; TotalTime = 34.2645s; SamplesPerSecond = 747.1
+Finished Epoch[49 of 160]: [Training Set] TrainLossPerSample = 0.19823113; EvalErrPerSample = 0.068580002; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9231
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.49'
+Starting Epoch 50: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[50 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.19133921; EvalErr[0]PerSample = 0.06726562; TotalTime = 34.2493s; SamplesPerSecond = 747.5
+Finished Epoch[50 of 160]: [Training Set] TrainLossPerSample = 0.2006802; EvalErrPerSample = 0.070699997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9319
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.50'
+Starting Epoch 51: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[51 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.19296997; EvalErr[0]PerSample = 0.06621094; TotalTime = 34.2297s; SamplesPerSecond = 747.9
+Finished Epoch[51 of 160]: [Training Set] TrainLossPerSample = 0.1992881; EvalErrPerSample = 0.069499999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8847
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.51'
+Starting Epoch 52: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[52 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18573498; EvalErr[0]PerSample = 0.06414063; TotalTime = 34.2456s; SamplesPerSecond = 747.5
+Finished Epoch[52 of 160]: [Training Set] TrainLossPerSample = 0.19471373; EvalErrPerSample = 0.067099996; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9424
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.52'
+Starting Epoch 53: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[53 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18396889; EvalErr[0]PerSample = 0.06542969; TotalTime = 34.2888s; SamplesPerSecond = 746.6
+Finished Epoch[53 of 160]: [Training Set] TrainLossPerSample = 0.1925908; EvalErrPerSample = 0.067219995; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9471
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.53'
+Starting Epoch 54: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[54 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18691126; EvalErr[0]PerSample = 0.06562500; TotalTime = 34.2456s; SamplesPerSecond = 747.5
+Finished Epoch[54 of 160]: [Training Set] TrainLossPerSample = 0.18951587; EvalErrPerSample = 0.065919995; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9116
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.54'
+Starting Epoch 55: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[55 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18601879; EvalErr[0]PerSample = 0.06371094; TotalTime = 34.2466s; SamplesPerSecond = 747.5
+Finished Epoch[55 of 160]: [Training Set] TrainLossPerSample = 0.18988976; EvalErrPerSample = 0.066100001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9152
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.55'
+Starting Epoch 56: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[56 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18315941; EvalErr[0]PerSample = 0.06406250; TotalTime = 34.2413s; SamplesPerSecond = 747.6
+Finished Epoch[56 of 160]: [Training Set] TrainLossPerSample = 0.18686067; EvalErrPerSample = 0.065739997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9069
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.56'
+Starting Epoch 57: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[57 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18106951; EvalErr[0]PerSample = 0.06246094; TotalTime = 34.2510s; SamplesPerSecond = 747.4
+Finished Epoch[57 of 160]: [Training Set] TrainLossPerSample = 0.19042736; EvalErrPerSample = 0.06622; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9223
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.57'
+Starting Epoch 58: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[58 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.18492083; EvalErr[0]PerSample = 0.06527344; TotalTime = 34.2528s; SamplesPerSecond = 747.4
+Finished Epoch[58 of 160]: [Training Set] TrainLossPerSample = 0.18727326; EvalErrPerSample = 0.064939998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9267
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.58'
+Starting Epoch 59: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[59 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17038435; EvalErr[0]PerSample = 0.05917969; TotalTime = 34.2496s; SamplesPerSecond = 747.5
+Finished Epoch[59 of 160]: [Training Set] TrainLossPerSample = 0.18080178; EvalErrPerSample = 0.062299997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9191
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.59'
+Starting Epoch 60: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[60 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17502851; EvalErr[0]PerSample = 0.06183594; TotalTime = 34.2793s; SamplesPerSecond = 746.8
+Finished Epoch[60 of 160]: [Training Set] TrainLossPerSample = 0.17897207; EvalErrPerSample = 0.062819995; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9611
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.60'
+Starting Epoch 61: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[61 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17479221; EvalErr[0]PerSample = 0.06187500; TotalTime = 34.2481s; SamplesPerSecond = 747.5
+Finished Epoch[61 of 160]: [Training Set] TrainLossPerSample = 0.17798932; EvalErrPerSample = 0.062660001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9219
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.61'
+Starting Epoch 62: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[62 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17981518; EvalErr[0]PerSample = 0.06199219; TotalTime = 34.2561s; SamplesPerSecond = 747.3
+Finished Epoch[62 of 160]: [Training Set] TrainLossPerSample = 0.18243115; EvalErrPerSample = 0.064240001; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9244
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.62'
+Starting Epoch 63: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[63 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17270412; EvalErr[0]PerSample = 0.05917969; TotalTime = 34.2554s; SamplesPerSecond = 747.3
+Finished Epoch[63 of 160]: [Training Set] TrainLossPerSample = 0.17758152; EvalErrPerSample = 0.061719999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9361
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.63'
+Starting Epoch 64: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[64 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16516579; EvalErr[0]PerSample = 0.05792969; TotalTime = 34.2766s; SamplesPerSecond = 746.9
+Finished Epoch[64 of 160]: [Training Set] TrainLossPerSample = 0.17424898; EvalErrPerSample = 0.06044; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9633
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.64'
+Starting Epoch 65: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[65 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.17415672; EvalErr[0]PerSample = 0.06164062; TotalTime = 34.3307s; SamplesPerSecond = 745.7
+Finished Epoch[65 of 160]: [Training Set] TrainLossPerSample = 0.17599966; EvalErrPerSample = 0.062039997; AvgLearningRatePerSample = 0.0078125; EpochTime=67.026
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.65'
+Starting Epoch 66: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[66 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16262278; EvalErr[0]PerSample = 0.05886719; TotalTime = 34.2448s; SamplesPerSecond = 747.6
+Finished Epoch[66 of 160]: [Training Set] TrainLossPerSample = 0.16772267; EvalErrPerSample = 0.059799999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9065
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.66'
+Starting Epoch 67: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[67 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16856216; EvalErr[0]PerSample = 0.05906250; TotalTime = 34.2467s; SamplesPerSecond = 747.5
+Finished Epoch[67 of 160]: [Training Set] TrainLossPerSample = 0.17300676; EvalErrPerSample = 0.061139997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9215
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.67'
+Starting Epoch 68: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[68 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16585386; EvalErr[0]PerSample = 0.05734375; TotalTime = 34.2464s; SamplesPerSecond = 747.5
+Finished Epoch[68 of 160]: [Training Set] TrainLossPerSample = 0.17578007; EvalErrPerSample = 0.060519997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9115
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.68'
+Starting Epoch 69: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[69 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15487971; EvalErr[0]PerSample = 0.05441406; TotalTime = 34.2490s; SamplesPerSecond = 747.5
+Finished Epoch[69 of 160]: [Training Set] TrainLossPerSample = 0.16411984; EvalErrPerSample = 0.05844; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9135
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.69'
+Starting Epoch 70: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[70 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15948469; EvalErr[0]PerSample = 0.05597656; TotalTime = 34.2446s; SamplesPerSecond = 747.6
+Finished Epoch[70 of 160]: [Training Set] TrainLossPerSample = 0.16847; EvalErrPerSample = 0.058879998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9103
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.70'
+Starting Epoch 71: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[71 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16197550; EvalErr[0]PerSample = 0.05718750; TotalTime = 34.2458s; SamplesPerSecond = 747.5
+Finished Epoch[71 of 160]: [Training Set] TrainLossPerSample = 0.17195615; EvalErrPerSample = 0.060839999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9166
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.71'
+Starting Epoch 72: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[72 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15629568; EvalErr[0]PerSample = 0.05515625; TotalTime = 34.2503s; SamplesPerSecond = 747.4
+Finished Epoch[72 of 160]: [Training Set] TrainLossPerSample = 0.17160599; EvalErrPerSample = 0.061219998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9158
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.72'
+Starting Epoch 73: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[73 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15722611; EvalErr[0]PerSample = 0.05476562; TotalTime = 34.2420s; SamplesPerSecond = 747.6
+Finished Epoch[73 of 160]: [Training Set] TrainLossPerSample = 0.16512014; EvalErrPerSample = 0.058259998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.8986
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.73'
+Starting Epoch 74: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[74 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.16009672; EvalErr[0]PerSample = 0.05703125; TotalTime = 34.2495s; SamplesPerSecond = 747.5
+Finished Epoch[74 of 160]: [Training Set] TrainLossPerSample = 0.16898924; EvalErrPerSample = 0.059019998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9233
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.74'
+Starting Epoch 75: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[75 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15460204; EvalErr[0]PerSample = 0.05312500; TotalTime = 34.2539s; SamplesPerSecond = 747.4
+Finished Epoch[75 of 160]: [Training Set] TrainLossPerSample = 0.16324201; EvalErrPerSample = 0.055919997; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9233
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.75'
+Starting Epoch 76: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[76 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15667317; EvalErr[0]PerSample = 0.05515625; TotalTime = 34.2399s; SamplesPerSecond = 747.7
+Finished Epoch[76 of 160]: [Training Set] TrainLossPerSample = 0.15858564; EvalErrPerSample = 0.055059999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9518
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.76'
+Starting Epoch 77: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[77 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15659092; EvalErr[0]PerSample = 0.05410156; TotalTime = 34.2332s; SamplesPerSecond = 747.8
+Finished Epoch[77 of 160]: [Training Set] TrainLossPerSample = 0.16720241; EvalErrPerSample = 0.058079999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9014
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.77'
+Starting Epoch 78: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[78 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15273868; EvalErr[0]PerSample = 0.05164063; TotalTime = 34.2404s; SamplesPerSecond = 747.7
+Finished Epoch[78 of 160]: [Training Set] TrainLossPerSample = 0.15992303; EvalErrPerSample = 0.055239998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.906
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.78'
+Starting Epoch 79: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[79 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15322983; EvalErr[0]PerSample = 0.05476562; TotalTime = 34.2420s; SamplesPerSecond = 747.6
+Finished Epoch[79 of 160]: [Training Set] TrainLossPerSample = 0.16380094; EvalErrPerSample = 0.058139998; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9021
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.79'
+Starting Epoch 80: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[80 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.15410002; EvalErr[0]PerSample = 0.05335938; TotalTime = 34.2510s; SamplesPerSecond = 747.4
+Finished Epoch[80 of 160]: [Training Set] TrainLossPerSample = 0.16132115; EvalErrPerSample = 0.05638; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9122
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.80'
+Starting Epoch 81: learning rate per sample = 0.007813  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[81 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.14658320; EvalErr[0]PerSample = 0.05156250; TotalTime = 34.2506s; SamplesPerSecond = 747.4
+Finished Epoch[81 of 160]: [Training Set] TrainLossPerSample = 0.15670048; EvalErrPerSample = 0.055119999; AvgLearningRatePerSample = 0.0078125; EpochTime=66.9019
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.81'
+Starting Epoch 82: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[82 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.11055239; EvalErr[0]PerSample = 0.03714844; TotalTime = 34.2453s; SamplesPerSecond = 747.5
+Finished Epoch[82 of 160]: [Training Set] TrainLossPerSample = 0.090582155; EvalErrPerSample = 0.03066; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.8937
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.82'
+Starting Epoch 83: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[83 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.05391692; EvalErr[0]PerSample = 0.01675781; TotalTime = 34.2464s; SamplesPerSecond = 747.5
+Finished Epoch[83 of 160]: [Training Set] TrainLossPerSample = 0.052728198; EvalErrPerSample = 0.01644; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9025
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.83'
+Starting Epoch 84: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[84 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.04134419; EvalErr[0]PerSample = 0.01242188; TotalTime = 34.2487s; SamplesPerSecond = 747.5
+Finished Epoch[84 of 160]: [Training Set] TrainLossPerSample = 0.04261167; EvalErrPerSample = 0.01278; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.922
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.84'
+Starting Epoch 85: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[85 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03696221; EvalErr[0]PerSample = 0.01214844; TotalTime = 34.2502s; SamplesPerSecond = 747.4
+Finished Epoch[85 of 160]: [Training Set] TrainLossPerSample = 0.035290603; EvalErrPerSample = 0.01108; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9713
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.85'
+Starting Epoch 86: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[86 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.03103620; EvalErr[0]PerSample = 0.00906250; TotalTime = 34.2539s; SamplesPerSecond = 747.4
+Finished Epoch[86 of 160]: [Training Set] TrainLossPerSample = 0.031053338; EvalErrPerSample = 0.0093200002; AvgLearningRatePerSample = 0.00078125001; EpochTime=67.0851
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.86'
+Starting Epoch 87: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[87 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02747848; EvalErr[0]PerSample = 0.00800781; TotalTime = 34.9682s; SamplesPerSecond = 732.1
+Finished Epoch[87 of 160]: [Training Set] TrainLossPerSample = 0.028250355; EvalErrPerSample = 0.00832; AvgLearningRatePerSample = 0.00078125001; EpochTime=67.6314
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.87'
+Starting Epoch 88: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[88 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02383066; EvalErr[0]PerSample = 0.00714844; TotalTime = 34.2475s; SamplesPerSecond = 747.5
+Finished Epoch[88 of 160]: [Training Set] TrainLossPerSample = 0.023733234; EvalErrPerSample = 0.0070599997; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9154
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.88'
+Starting Epoch 89: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[89 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02334737; EvalErr[0]PerSample = 0.00632813; TotalTime = 34.2406s; SamplesPerSecond = 747.7
+Finished Epoch[89 of 160]: [Training Set] TrainLossPerSample = 0.022844482; EvalErrPerSample = 0.0063999998; AvgLearningRatePerSample = 0.00078125001; EpochTime=67.0741
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.89'
+Starting Epoch 90: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[90 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02094710; EvalErr[0]PerSample = 0.00589844; TotalTime = 34.2465s; SamplesPerSecond = 747.5
+Finished Epoch[90 of 160]: [Training Set] TrainLossPerSample = 0.020934064; EvalErrPerSample = 0.0059599997; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9229
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.90'
+Starting Epoch 91: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[91 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.02006884; EvalErr[0]PerSample = 0.00628906; TotalTime = 34.2548s; SamplesPerSecond = 747.3
+Finished Epoch[91 of 160]: [Training Set] TrainLossPerSample = 0.019843206; EvalErrPerSample = 0.0061399997; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.914
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.91'
+Starting Epoch 92: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[92 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01826200; EvalErr[0]PerSample = 0.00542969; TotalTime = 34.2810s; SamplesPerSecond = 746.8
+Finished Epoch[92 of 160]: [Training Set] TrainLossPerSample = 0.018547757; EvalErrPerSample = 0.00538; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9512
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.92'
+Starting Epoch 93: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[93 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01654859; EvalErr[0]PerSample = 0.00453125; TotalTime = 34.2943s; SamplesPerSecond = 746.5
+Finished Epoch[93 of 160]: [Training Set] TrainLossPerSample = 0.01722255; EvalErrPerSample = 0.0049399999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9701
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.93'
+Starting Epoch 94: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[94 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01585697; EvalErr[0]PerSample = 0.00441406; TotalTime = 34.2504s; SamplesPerSecond = 747.4
+Finished Epoch[94 of 160]: [Training Set] TrainLossPerSample = 0.01515844; EvalErrPerSample = 0.0041799997; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.93
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.94'
+Starting Epoch 95: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[95 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01532144; EvalErr[0]PerSample = 0.00441406; TotalTime = 34.2512s; SamplesPerSecond = 747.4
+Finished Epoch[95 of 160]: [Training Set] TrainLossPerSample = 0.015015967; EvalErrPerSample = 0.0044; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9121
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.95'
+Starting Epoch 96: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[96 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01399448; EvalErr[0]PerSample = 0.00406250; TotalTime = 34.2433s; SamplesPerSecond = 747.6
+Finished Epoch[96 of 160]: [Training Set] TrainLossPerSample = 0.013210027; EvalErrPerSample = 0.00376; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.8997
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.96'
+Starting Epoch 97: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[97 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01209825; EvalErr[0]PerSample = 0.00332031; TotalTime = 34.2567s; SamplesPerSecond = 747.3
+Finished Epoch[97 of 160]: [Training Set] TrainLossPerSample = 0.012226926; EvalErrPerSample = 0.0032599999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9247
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.97'
+Starting Epoch 98: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[98 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01230964; EvalErr[0]PerSample = 0.00343750; TotalTime = 34.2486s; SamplesPerSecond = 747.5
+Finished Epoch[98 of 160]: [Training Set] TrainLossPerSample = 0.012122833; EvalErrPerSample = 0.0034599998; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9192
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.98'
+Starting Epoch 99: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[99 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00988823; EvalErr[0]PerSample = 0.00230469; TotalTime = 34.2477s; SamplesPerSecond = 747.5
+Finished Epoch[99 of 160]: [Training Set] TrainLossPerSample = 0.011153033; EvalErrPerSample = 0.00288; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9079
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.99'
+Starting Epoch 100: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[100 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01078802; EvalErr[0]PerSample = 0.00328125; TotalTime = 34.2446s; SamplesPerSecond = 747.6
+Finished Epoch[100 of 160]: [Training Set] TrainLossPerSample = 0.010805478; EvalErrPerSample = 0.0032199998; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9141
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.100'
+Starting Epoch 101: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[101 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01043408; EvalErr[0]PerSample = 0.00285156; TotalTime = 34.2448s; SamplesPerSecond = 747.6
+Finished Epoch[101 of 160]: [Training Set] TrainLossPerSample = 0.0097289206; EvalErrPerSample = 0.00254; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9115
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.101'
+Starting Epoch 102: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[102 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01028690; EvalErr[0]PerSample = 0.00269531; TotalTime = 34.2500s; SamplesPerSecond = 747.4
+Finished Epoch[102 of 160]: [Training Set] TrainLossPerSample = 0.010017196; EvalErrPerSample = 0.0026999998; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9075
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.102'
+Starting Epoch 103: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[103 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.01031995; EvalErr[0]PerSample = 0.00316406; TotalTime = 34.2485s; SamplesPerSecond = 747.5
+Finished Epoch[103 of 160]: [Training Set] TrainLossPerSample = 0.010233736; EvalErrPerSample = 0.0030199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9094
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.103'
+Starting Epoch 104: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[104 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00783067; EvalErr[0]PerSample = 0.00164062; TotalTime = 34.2752s; SamplesPerSecond = 746.9
+Finished Epoch[104 of 160]: [Training Set] TrainLossPerSample = 0.0086551448; EvalErrPerSample = 0.00208; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9965
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.104'
+Starting Epoch 105: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[105 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00939923; EvalErr[0]PerSample = 0.00253906; TotalTime = 34.2407s; SamplesPerSecond = 747.6
+Finished Epoch[105 of 160]: [Training Set] TrainLossPerSample = 0.0096224733; EvalErrPerSample = 0.0026199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.8974
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.105'
+Starting Epoch 106: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[106 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00753223; EvalErr[0]PerSample = 0.00179688; TotalTime = 34.2848s; SamplesPerSecond = 746.7
+Finished Epoch[106 of 160]: [Training Set] TrainLossPerSample = 0.0078934198; EvalErrPerSample = 0.00202; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9655
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.106'
+Starting Epoch 107: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[107 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00865439; EvalErr[0]PerSample = 0.00238281; TotalTime = 34.2544s; SamplesPerSecond = 747.4
+Finished Epoch[107 of 160]: [Training Set] TrainLossPerSample = 0.008151643; EvalErrPerSample = 0.00214; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9284
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.107'
+Starting Epoch 108: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[108 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00694811; EvalErr[0]PerSample = 0.00148438; TotalTime = 34.2490s; SamplesPerSecond = 747.5
+Finished Epoch[108 of 160]: [Training Set] TrainLossPerSample = 0.0073725204; EvalErrPerSample = 0.0016; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9107
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.108'
+Starting Epoch 109: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[109 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00807542; EvalErr[0]PerSample = 0.00218750; TotalTime = 34.2420s; SamplesPerSecond = 747.6
+Finished Epoch[109 of 160]: [Training Set] TrainLossPerSample = 0.0075115636; EvalErrPerSample = 0.00194; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9172
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.109'
+Starting Epoch 110: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[110 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00747269; EvalErr[0]PerSample = 0.00214844; TotalTime = 34.2519s; SamplesPerSecond = 747.4
+Finished Epoch[110 of 160]: [Training Set] TrainLossPerSample = 0.0070967474; EvalErrPerSample = 0.0017599999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9183
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.110'
+Starting Epoch 111: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[111 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00696611; EvalErr[0]PerSample = 0.00175781; TotalTime = 34.2387s; SamplesPerSecond = 747.7
+Finished Epoch[111 of 160]: [Training Set] TrainLossPerSample = 0.0071680127; EvalErrPerSample = 0.0018399999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9065
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.111'
+Starting Epoch 112: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[112 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00660637; EvalErr[0]PerSample = 0.00167969; TotalTime = 34.2517s; SamplesPerSecond = 747.4
+Finished Epoch[112 of 160]: [Training Set] TrainLossPerSample = 0.0069257859; EvalErrPerSample = 0.0018399999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9221
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.112'
+Starting Epoch 113: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[113 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00683305; EvalErr[0]PerSample = 0.00179688; TotalTime = 34.2447s; SamplesPerSecond = 747.6
+Finished Epoch[113 of 160]: [Training Set] TrainLossPerSample = 0.0071453024; EvalErrPerSample = 0.00182; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9078
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.113'
+Starting Epoch 114: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[114 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00659715; EvalErr[0]PerSample = 0.00164062; TotalTime = 34.2397s; SamplesPerSecond = 747.7
+Finished Epoch[114 of 160]: [Training Set] TrainLossPerSample = 0.0066223624; EvalErrPerSample = 0.0017199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9022
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.114'
+Starting Epoch 115: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[115 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00571319; EvalErr[0]PerSample = 0.00140625; TotalTime = 34.2379s; SamplesPerSecond = 747.7
+Finished Epoch[115 of 160]: [Training Set] TrainLossPerSample = 0.0061274339; EvalErrPerSample = 0.0015199999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9117
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.115'
+Starting Epoch 116: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[116 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00616163; EvalErr[0]PerSample = 0.00164062; TotalTime = 34.2441s; SamplesPerSecond = 747.6
+Finished Epoch[116 of 160]: [Training Set] TrainLossPerSample = 0.0064637884; EvalErrPerSample = 0.00166; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.92
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.116'
+Starting Epoch 117: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[117 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00667505; EvalErr[0]PerSample = 0.00171875; TotalTime = 34.2353s; SamplesPerSecond = 747.8
+Finished Epoch[117 of 160]: [Training Set] TrainLossPerSample = 0.0063761352; EvalErrPerSample = 0.0016; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.8921
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.117'
+Starting Epoch 118: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[118 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00584337; EvalErr[0]PerSample = 0.00144531; TotalTime = 34.2344s; SamplesPerSecond = 747.8
+Finished Epoch[118 of 160]: [Training Set] TrainLossPerSample = 0.0061604949; EvalErrPerSample = 0.00162; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9005
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.118'
+Starting Epoch 119: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[119 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00587101; EvalErr[0]PerSample = 0.00132813; TotalTime = 34.2448s; SamplesPerSecond = 747.6
+Finished Epoch[119 of 160]: [Training Set] TrainLossPerSample = 0.0058231312; EvalErrPerSample = 0.0013799999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9065
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.119'
+Starting Epoch 120: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[120 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00556109; EvalErr[0]PerSample = 0.00152344; TotalTime = 34.2181s; SamplesPerSecond = 748.1
+Finished Epoch[120 of 160]: [Training Set] TrainLossPerSample = 0.0049992148; EvalErrPerSample = 0.0013; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9296
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.120'
+Starting Epoch 121: learning rate per sample = 0.000781  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[121 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00544529; EvalErr[0]PerSample = 0.00148438; TotalTime = 34.2507s; SamplesPerSecond = 747.4
+Finished Epoch[121 of 160]: [Training Set] TrainLossPerSample = 0.0059123267; EvalErrPerSample = 0.0016999999; AvgLearningRatePerSample = 0.00078125001; EpochTime=66.9099
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.121'
+Starting Epoch 122: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[122 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00506212; EvalErr[0]PerSample = 0.00109375; TotalTime = 34.2380s; SamplesPerSecond = 747.7
+Finished Epoch[122 of 160]: [Training Set] TrainLossPerSample = 0.0047056451; EvalErrPerSample = 0.00099999993; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8882
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.122'
+Starting Epoch 123: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[123 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00436568; EvalErr[0]PerSample = 0.00093750; TotalTime = 34.2202s; SamplesPerSecond = 748.1
+Finished Epoch[123 of 160]: [Training Set] TrainLossPerSample = 0.0045333277; EvalErrPerSample = 0.00102; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8649
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.123'
+Starting Epoch 124: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[124 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00423286; EvalErr[0]PerSample = 0.00089844; TotalTime = 34.2368s; SamplesPerSecond = 747.7
+Finished Epoch[124 of 160]: [Training Set] TrainLossPerSample = 0.0040925727; EvalErrPerSample = 0.00078; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8901
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.124'
+Starting Epoch 125: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[125 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00428819; EvalErr[0]PerSample = 0.00089844; TotalTime = 34.2302s; SamplesPerSecond = 747.9
+Finished Epoch[125 of 160]: [Training Set] TrainLossPerSample = 0.0039790319; EvalErrPerSample = 0.00083999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8868
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.125'
+Starting Epoch 126: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[126 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00386990; EvalErr[0]PerSample = 0.00070312; TotalTime = 34.2288s; SamplesPerSecond = 747.9
+Finished Epoch[126 of 160]: [Training Set] TrainLossPerSample = 0.0041431645; EvalErrPerSample = 0.00102; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8935
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.126'
+Starting Epoch 127: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[127 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00415913; EvalErr[0]PerSample = 0.00074219; TotalTime = 34.2434s; SamplesPerSecond = 747.6
+Finished Epoch[127 of 160]: [Training Set] TrainLossPerSample = 0.0038879183; EvalErrPerSample = 0.00072000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9178
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.127'
+Starting Epoch 128: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[128 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00381589; EvalErr[0]PerSample = 0.00074219; TotalTime = 34.2352s; SamplesPerSecond = 747.8
+Finished Epoch[128 of 160]: [Training Set] TrainLossPerSample = 0.0037014519; EvalErrPerSample = 0.00065999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8974
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.128'
+Starting Epoch 129: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[129 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00390699; EvalErr[0]PerSample = 0.00093750; TotalTime = 34.2443s; SamplesPerSecond = 747.6
+Finished Epoch[129 of 160]: [Training Set] TrainLossPerSample = 0.0036388342; EvalErrPerSample = 0.00073999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9066
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.129'
+Starting Epoch 130: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[130 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00401969; EvalErr[0]PerSample = 0.00113281; TotalTime = 34.2339s; SamplesPerSecond = 747.8
+Finished Epoch[130 of 160]: [Training Set] TrainLossPerSample = 0.0038497401; EvalErrPerSample = 0.00089999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8803
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.130'
+Starting Epoch 131: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[131 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00371109; EvalErr[0]PerSample = 0.00050781; TotalTime = 34.2393s; SamplesPerSecond = 747.7
+Finished Epoch[131 of 160]: [Training Set] TrainLossPerSample = 0.0036305008; EvalErrPerSample = 0.00055999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9068
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.131'
+Starting Epoch 132: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[132 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00361982; EvalErr[0]PerSample = 0.00082031; TotalTime = 34.2404s; SamplesPerSecond = 747.7
+Finished Epoch[132 of 160]: [Training Set] TrainLossPerSample = 0.0037875406; EvalErrPerSample = 0.00091999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9025
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.132'
+Starting Epoch 133: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[133 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00379993; EvalErr[0]PerSample = 0.00085938; TotalTime = 34.2434s; SamplesPerSecond = 747.6
+Finished Epoch[133 of 160]: [Training Set] TrainLossPerSample = 0.0037811422; EvalErrPerSample = 0.00072000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8997
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.133'
+Starting Epoch 134: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[134 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00324624; EvalErr[0]PerSample = 0.00042969; TotalTime = 34.2358s; SamplesPerSecond = 747.8
+Finished Epoch[134 of 160]: [Training Set] TrainLossPerSample = 0.0031919111; EvalErrPerSample = 0.00052; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8704
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.134'
+Starting Epoch 135: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[135 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00297883; EvalErr[0]PerSample = 0.00035156; TotalTime = 34.2219s; SamplesPerSecond = 748.1
+Finished Epoch[135 of 160]: [Training Set] TrainLossPerSample = 0.0031994823; EvalErrPerSample = 0.00059999997; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9955
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.135'
+Starting Epoch 136: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[136 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00355394; EvalErr[0]PerSample = 0.00066406; TotalTime = 34.9371s; SamplesPerSecond = 732.7
+Finished Epoch[136 of 160]: [Training Set] TrainLossPerSample = 0.0036795642; EvalErrPerSample = 0.00078; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=67.6302
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.136'
+Starting Epoch 137: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[137 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00359524; EvalErr[0]PerSample = 0.00066406; TotalTime = 34.2354s; SamplesPerSecond = 747.8
+Finished Epoch[137 of 160]: [Training Set] TrainLossPerSample = 0.0035311766; EvalErrPerSample = 0.00068; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.891
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.137'
+Starting Epoch 138: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[138 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00343087; EvalErr[0]PerSample = 0.00066406; TotalTime = 34.2621s; SamplesPerSecond = 747.2
+Finished Epoch[138 of 160]: [Training Set] TrainLossPerSample = 0.0034796586; EvalErrPerSample = 0.00065999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.916
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.138'
+Starting Epoch 139: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[139 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00343002; EvalErr[0]PerSample = 0.00066406; TotalTime = 34.2436s; SamplesPerSecond = 747.6
+Finished Epoch[139 of 160]: [Training Set] TrainLossPerSample = 0.0033328664; EvalErrPerSample = 0.00055999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.923
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.139'
+Starting Epoch 140: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[140 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00337591; EvalErr[0]PerSample = 0.00070312; TotalTime = 34.2353s; SamplesPerSecond = 747.8
+Finished Epoch[140 of 160]: [Training Set] TrainLossPerSample = 0.0031642318; EvalErrPerSample = 0.00055999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.926
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.140'
+Starting Epoch 141: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[141 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00364618; EvalErr[0]PerSample = 0.00078125; TotalTime = 34.2214s; SamplesPerSecond = 748.1
+Finished Epoch[141 of 160]: [Training Set] TrainLossPerSample = 0.0035639035; EvalErrPerSample = 0.00078; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8739
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.141'
+Starting Epoch 142: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[142 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00340368; EvalErr[0]PerSample = 0.00046875; TotalTime = 34.2602s; SamplesPerSecond = 747.2
+Finished Epoch[142 of 160]: [Training Set] TrainLossPerSample = 0.0033217999; EvalErrPerSample = 0.00043999997; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9173
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.142'
+Starting Epoch 143: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[143 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00338313; EvalErr[0]PerSample = 0.00058594; TotalTime = 34.2287s; SamplesPerSecond = 747.9
+Finished Epoch[143 of 160]: [Training Set] TrainLossPerSample = 0.003285937; EvalErrPerSample = 0.00055999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8929
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.143'
+Starting Epoch 144: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[144 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00350241; EvalErr[0]PerSample = 0.00097656; TotalTime = 34.2369s; SamplesPerSecond = 747.7
+Finished Epoch[144 of 160]: [Training Set] TrainLossPerSample = 0.0033956808; EvalErrPerSample = 0.00087999995; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9109
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.144'
+Starting Epoch 145: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[145 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00389780; EvalErr[0]PerSample = 0.00078125; TotalTime = 34.2499s; SamplesPerSecond = 747.4
+Finished Epoch[145 of 160]: [Training Set] TrainLossPerSample = 0.003525309; EvalErrPerSample = 0.00073999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9108
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.145'
+Starting Epoch 146: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[146 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00273976; EvalErr[0]PerSample = 0.00050781; TotalTime = 34.2481s; SamplesPerSecond = 747.5
+Finished Epoch[146 of 160]: [Training Set] TrainLossPerSample = 0.0029731863; EvalErrPerSample = 0.00049999997; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9057
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.146'
+Starting Epoch 147: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[147 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00301678; EvalErr[0]PerSample = 0.00054688; TotalTime = 34.2455s; SamplesPerSecond = 747.5
+Finished Epoch[147 of 160]: [Training Set] TrainLossPerSample = 0.0028605016; EvalErrPerSample = 0.00055999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9068
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.147'
+Starting Epoch 148: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[148 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00318503; EvalErr[0]PerSample = 0.00078125; TotalTime = 34.2349s; SamplesPerSecond = 747.8
+Finished Epoch[148 of 160]: [Training Set] TrainLossPerSample = 0.0029674848; EvalErrPerSample = 0.00063999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8889
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.148'
+Starting Epoch 149: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[149 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00341538; EvalErr[0]PerSample = 0.00070312; TotalTime = 34.2300s; SamplesPerSecond = 747.9
+Finished Epoch[149 of 160]: [Training Set] TrainLossPerSample = 0.0034138309; EvalErrPerSample = 0.00062000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.973
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.149'
+Starting Epoch 150: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[150 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00312641; EvalErr[0]PerSample = 0.00050781; TotalTime = 34.2394s; SamplesPerSecond = 747.7
+Finished Epoch[150 of 160]: [Training Set] TrainLossPerSample = 0.0032544835; EvalErrPerSample = 0.00062000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8886
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.150'
+Starting Epoch 151: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[151 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00294002; EvalErr[0]PerSample = 0.00046875; TotalTime = 34.2440s; SamplesPerSecond = 747.6
+Finished Epoch[151 of 160]: [Training Set] TrainLossPerSample = 0.0029587755; EvalErrPerSample = 0.00049999997; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9165
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.151'
+Starting Epoch 152: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[152 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00312315; EvalErr[0]PerSample = 0.00070312; TotalTime = 34.2407s; SamplesPerSecond = 747.6
+Finished Epoch[152 of 160]: [Training Set] TrainLossPerSample = 0.0031671789; EvalErrPerSample = 0.00063999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8987
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.152'
+Starting Epoch 153: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[153 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00293371; EvalErr[0]PerSample = 0.00042969; TotalTime = 34.2360s; SamplesPerSecond = 747.8
+Finished Epoch[153 of 160]: [Training Set] TrainLossPerSample = 0.0029186283; EvalErrPerSample = 0.00036000001; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8934
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.153'
+Starting Epoch 154: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[154 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00270366; EvalErr[0]PerSample = 0.00046875; TotalTime = 34.2303s; SamplesPerSecond = 747.9
+Finished Epoch[154 of 160]: [Training Set] TrainLossPerSample = 0.0027337885; EvalErrPerSample = 0.00029999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8932
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.154'
+Starting Epoch 155: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[155 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00262628; EvalErr[0]PerSample = 0.00027344; TotalTime = 34.2340s; SamplesPerSecond = 747.8
+Finished Epoch[155 of 160]: [Training Set] TrainLossPerSample = 0.002633905; EvalErrPerSample = 0.00023999999; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8889
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.155'
+Starting Epoch 156: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[156 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00307139; EvalErr[0]PerSample = 0.00074219; TotalTime = 34.2388s; SamplesPerSecond = 747.7
+Finished Epoch[156 of 160]: [Training Set] TrainLossPerSample = 0.0029770101; EvalErrPerSample = 0.00065999996; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8912
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.156'
+Starting Epoch 157: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[157 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00275711; EvalErr[0]PerSample = 0.00031250; TotalTime = 34.2329s; SamplesPerSecond = 747.8
+Finished Epoch[157 of 160]: [Training Set] TrainLossPerSample = 0.0026996653; EvalErrPerSample = 0.00034; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8768
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.157'
+Starting Epoch 158: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[158 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00278720; EvalErr[0]PerSample = 0.00042969; TotalTime = 34.2421s; SamplesPerSecond = 747.6
+Finished Epoch[158 of 160]: [Training Set] TrainLossPerSample = 0.0026968825; EvalErrPerSample = 0.00037999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9344
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.158'
+Starting Epoch 159: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[159 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00272500; EvalErr[0]PerSample = 0.00031250; TotalTime = 34.2259s; SamplesPerSecond = 748.0
+Finished Epoch[159 of 160]: [Training Set] TrainLossPerSample = 0.0027625638; EvalErrPerSample = 0.00037999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.9162
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56.159'
+Starting Epoch 160: learning rate per sample = 0.000078  effective momentum = 0.900000  momentum as time constant = 1214.9 samples
+
+Starting minibatch loop.
+ Epoch[160 of 160]-Minibatch[   1- 200, 100.00%]: SamplesSeen = 25600; TrainLossPerSample =  0.00257069; EvalErr[0]PerSample = 0.00042969; TotalTime = 34.2364s; SamplesPerSecond = 747.7
+Finished Epoch[160 of 160]: [Training Set] TrainLossPerSample = 0.0029211191; EvalErrPerSample = 0.00063999998; AvgLearningRatePerSample = 7.8124998e-005; EpochTime=66.8957
+SGD: Saving checkpoint model './Output/Models/04_ResNet_56'
+CNTKCommandTrainEnd: Train
+
+Post-processing network...
+
+3 roots:
+	CE = CrossEntropyWithSoftmax
+	Err = ErrorPrediction
+	OutputNodes.z = Plus
+FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
+FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
+FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
+
+
+Validating network. 949 nodes to process in pass 1.
+
+Validating --> labels = InputValue -> [10 x *]
+Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
+Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
+Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
+Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
+Validating --> features = InputValue -> [32 x 32 x 3 x *]
+Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
+Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
+Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
+Validating --> OutputNodes.b = LearnableParameter -> [10]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+
+Validating network. 390 nodes to process in pass 2.
+
+Validating --> labels = InputValue -> [10 x *]
+Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
+Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
+Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
+Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
+Validating --> features = InputValue -> [32 x 32 x 3 x *]
+Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
+Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
+Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
+Validating --> OutputNodes.b = LearnableParameter -> [10]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+
+Validating network, final pass.
+
+Validating --> labels = InputValue -> [10 x *]
+Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
+Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
+Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
+Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
+Validating --> features = InputValue -> [32 x 32 x 3 x *]
+Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
+Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
+Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
+Validating --> OutputNodes.b = LearnableParameter -> [10]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+
+559 out of 949 nodes do not share the minibatch layout with the input data.
+
+Post-processing network complete.
+
+Post-processing network...
+
+3 roots:
+	CE = CrossEntropyWithSoftmax
+	Err = ErrorPrediction
+	OutputNodes.z = Plus
+FormNestedNetwork: WARNING: Was called twice for CE CrossEntropyWithSoftmax operation
+FormNestedNetwork: WARNING: Was called twice for Err ErrorPrediction operation
+FormNestedNetwork: WARNING: Was called twice for OutputNodes.z Plus operation
+
+
+Validating network. 949 nodes to process in pass 1.
+
+Validating --> labels = InputValue -> [10 x *]
+Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
+Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
+Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
+Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
+Validating --> features = InputValue -> [32 x 32 x 3 x *]
+Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
+Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
+Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
+Validating --> OutputNodes.b = LearnableParameter -> [10]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+
+Validating network. 390 nodes to process in pass 2.
+
+Validating --> labels = InputValue -> [10 x *]
+Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
+Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
+Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
+Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
+Validating --> features = InputValue -> [32 x 32 x 3 x *]
+Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
+Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
+Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
+Validating --> OutputNodes.b = LearnableParameter -> [10]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+
+Validating network, final pass.
+
+Validating --> labels = InputValue -> [10 x *]
+Validating --> OutputNodes.W = LearnableParameter -> [10 x 64]
+Validating --> rn3_18.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_18.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_17.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_16.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_15.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_14.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_13.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_12.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_11.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_10.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_9.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_8.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_7.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_6.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_5.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_4.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_3.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_2.c1.c.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c2.W = LearnableParameter -> [64 x 576]
+Validating --> rn3_1.c1.c.W = LearnableParameter -> [64 x 288]
+Validating --> rn2_18.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_18.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_17.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_16.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_15.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_14.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_13.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_12.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_11.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_10.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_9.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_8.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_7.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_6.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_5.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_4.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_3.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_2.c1.c.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c2.W = LearnableParameter -> [32 x 288]
+Validating --> rn2_1.c1.c.W = LearnableParameter -> [32 x 144]
+Validating --> rn1_18.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_18.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_17.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_16.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_15.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_14.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_13.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_12.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_11.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_10.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_9.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_8.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_7.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_6.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_5.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_4.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_3.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_2.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c2.W = LearnableParameter -> [16 x 144]
+Validating --> rn1_1.c1.c.W = LearnableParameter -> [16 x 144]
+Validating --> conv1.c.W = LearnableParameter -> [16 x 27]
+Validating --> features = InputValue -> [32 x 32 x 3 x *]
+Validating --> conv1.c.c.c = Convolution(conv1.c.W[16 x 27], features[32 x 32 x 3 x * {W=32, H=3, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> conv1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> conv1.c.c.y = BatchNormalization(conv1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.c.c.sc[16 x 1], conv1.c.c.b[16 x 1], conv1.c.c.m[16 x 1], conv1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> conv1.y = RectifiedLinear(conv1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.c = Convolution(rn1_1.c1.c.W[16 x 144], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c1.c.c.y = BatchNormalization(rn1_1.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c1.c.c.sc[16 x 1], rn1_1.c1.c.c.b[16 x 1], rn1_1.c1.c.c.m[16 x 1], rn1_1.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c1.y = RectifiedLinear(rn1_1.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.c = Convolution(rn1_1.c2.W[16 x 144], rn1_1.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_1.c2.c.y = BatchNormalization(rn1_1.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.c2.c.sc[16 x 1], rn1_1.c2.c.b[16 x 1], rn1_1.c2.c.m[16 x 1], rn1_1.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.p = Plus(rn1_1.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], conv1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_1.y = RectifiedLinear(rn1_1.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.c = Convolution(rn1_2.c1.c.W[16 x 144], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c1.c.c.y = BatchNormalization(rn1_2.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c1.c.c.sc[16 x 1], rn1_2.c1.c.c.b[16 x 1], rn1_2.c1.c.c.m[16 x 1], rn1_2.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c1.y = RectifiedLinear(rn1_2.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.c = Convolution(rn1_2.c2.W[16 x 144], rn1_2.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_2.c2.c.y = BatchNormalization(rn1_2.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.c2.c.sc[16 x 1], rn1_2.c2.c.b[16 x 1], rn1_2.c2.c.m[16 x 1], rn1_2.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.p = Plus(rn1_2.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_2.y = RectifiedLinear(rn1_2.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.c = Convolution(rn1_3.c1.c.W[16 x 144], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c1.c.c.y = BatchNormalization(rn1_3.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c1.c.c.sc[16 x 1], rn1_3.c1.c.c.b[16 x 1], rn1_3.c1.c.c.m[16 x 1], rn1_3.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c1.y = RectifiedLinear(rn1_3.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.c = Convolution(rn1_3.c2.W[16 x 144], rn1_3.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_3.c2.c.y = BatchNormalization(rn1_3.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.c2.c.sc[16 x 1], rn1_3.c2.c.b[16 x 1], rn1_3.c2.c.m[16 x 1], rn1_3.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.p = Plus(rn1_3.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_2.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_3.y = RectifiedLinear(rn1_3.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.c = Convolution(rn1_4.c1.c.W[16 x 144], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c1.c.c.y = BatchNormalization(rn1_4.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c1.c.c.sc[16 x 1], rn1_4.c1.c.c.b[16 x 1], rn1_4.c1.c.c.m[16 x 1], rn1_4.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c1.y = RectifiedLinear(rn1_4.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.c = Convolution(rn1_4.c2.W[16 x 144], rn1_4.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_4.c2.c.y = BatchNormalization(rn1_4.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.c2.c.sc[16 x 1], rn1_4.c2.c.b[16 x 1], rn1_4.c2.c.m[16 x 1], rn1_4.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.p = Plus(rn1_4.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_3.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_4.y = RectifiedLinear(rn1_4.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.c = Convolution(rn1_5.c1.c.W[16 x 144], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c1.c.c.y = BatchNormalization(rn1_5.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c1.c.c.sc[16 x 1], rn1_5.c1.c.c.b[16 x 1], rn1_5.c1.c.c.m[16 x 1], rn1_5.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c1.y = RectifiedLinear(rn1_5.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.c = Convolution(rn1_5.c2.W[16 x 144], rn1_5.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_5.c2.c.y = BatchNormalization(rn1_5.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.c2.c.sc[16 x 1], rn1_5.c2.c.b[16 x 1], rn1_5.c2.c.m[16 x 1], rn1_5.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.p = Plus(rn1_5.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_4.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_5.y = RectifiedLinear(rn1_5.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.c = Convolution(rn1_6.c1.c.W[16 x 144], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c1.c.c.y = BatchNormalization(rn1_6.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c1.c.c.sc[16 x 1], rn1_6.c1.c.c.b[16 x 1], rn1_6.c1.c.c.m[16 x 1], rn1_6.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c1.y = RectifiedLinear(rn1_6.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.c = Convolution(rn1_6.c2.W[16 x 144], rn1_6.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_6.c2.c.y = BatchNormalization(rn1_6.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.c2.c.sc[16 x 1], rn1_6.c2.c.b[16 x 1], rn1_6.c2.c.m[16 x 1], rn1_6.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.p = Plus(rn1_6.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_5.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_6.y = RectifiedLinear(rn1_6.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.c = Convolution(rn1_7.c1.c.W[16 x 144], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c1.c.c.y = BatchNormalization(rn1_7.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c1.c.c.sc[16 x 1], rn1_7.c1.c.c.b[16 x 1], rn1_7.c1.c.c.m[16 x 1], rn1_7.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c1.y = RectifiedLinear(rn1_7.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.c = Convolution(rn1_7.c2.W[16 x 144], rn1_7.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_7.c2.c.y = BatchNormalization(rn1_7.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.c2.c.sc[16 x 1], rn1_7.c2.c.b[16 x 1], rn1_7.c2.c.m[16 x 1], rn1_7.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.p = Plus(rn1_7.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_6.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_7.y = RectifiedLinear(rn1_7.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.c = Convolution(rn1_8.c1.c.W[16 x 144], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c1.c.c.y = BatchNormalization(rn1_8.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c1.c.c.sc[16 x 1], rn1_8.c1.c.c.b[16 x 1], rn1_8.c1.c.c.m[16 x 1], rn1_8.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c1.y = RectifiedLinear(rn1_8.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.c = Convolution(rn1_8.c2.W[16 x 144], rn1_8.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_8.c2.c.y = BatchNormalization(rn1_8.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.c2.c.sc[16 x 1], rn1_8.c2.c.b[16 x 1], rn1_8.c2.c.m[16 x 1], rn1_8.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.p = Plus(rn1_8.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_7.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_8.y = RectifiedLinear(rn1_8.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.c = Convolution(rn1_9.c1.c.W[16 x 144], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c1.c.c.y = BatchNormalization(rn1_9.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c1.c.c.sc[16 x 1], rn1_9.c1.c.c.b[16 x 1], rn1_9.c1.c.c.m[16 x 1], rn1_9.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c1.y = RectifiedLinear(rn1_9.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.c = Convolution(rn1_9.c2.W[16 x 144], rn1_9.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_9.c2.c.y = BatchNormalization(rn1_9.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.c2.c.sc[16 x 1], rn1_9.c2.c.b[16 x 1], rn1_9.c2.c.m[16 x 1], rn1_9.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.p = Plus(rn1_9.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_8.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_9.y = RectifiedLinear(rn1_9.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.c = Convolution(rn1_10.c1.c.W[16 x 144], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c1.c.c.y = BatchNormalization(rn1_10.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c1.c.c.sc[16 x 1], rn1_10.c1.c.c.b[16 x 1], rn1_10.c1.c.c.m[16 x 1], rn1_10.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c1.y = RectifiedLinear(rn1_10.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.c = Convolution(rn1_10.c2.W[16 x 144], rn1_10.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_10.c2.c.y = BatchNormalization(rn1_10.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.c2.c.sc[16 x 1], rn1_10.c2.c.b[16 x 1], rn1_10.c2.c.m[16 x 1], rn1_10.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.p = Plus(rn1_10.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_9.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_10.y = RectifiedLinear(rn1_10.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.c = Convolution(rn1_11.c1.c.W[16 x 144], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c1.c.c.y = BatchNormalization(rn1_11.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c1.c.c.sc[16 x 1], rn1_11.c1.c.c.b[16 x 1], rn1_11.c1.c.c.m[16 x 1], rn1_11.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c1.y = RectifiedLinear(rn1_11.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.c = Convolution(rn1_11.c2.W[16 x 144], rn1_11.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_11.c2.c.y = BatchNormalization(rn1_11.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.c2.c.sc[16 x 1], rn1_11.c2.c.b[16 x 1], rn1_11.c2.c.m[16 x 1], rn1_11.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.p = Plus(rn1_11.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_10.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_11.y = RectifiedLinear(rn1_11.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.c = Convolution(rn1_12.c1.c.W[16 x 144], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c1.c.c.y = BatchNormalization(rn1_12.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c1.c.c.sc[16 x 1], rn1_12.c1.c.c.b[16 x 1], rn1_12.c1.c.c.m[16 x 1], rn1_12.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c1.y = RectifiedLinear(rn1_12.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.c = Convolution(rn1_12.c2.W[16 x 144], rn1_12.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_12.c2.c.y = BatchNormalization(rn1_12.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.c2.c.sc[16 x 1], rn1_12.c2.c.b[16 x 1], rn1_12.c2.c.m[16 x 1], rn1_12.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.p = Plus(rn1_12.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_11.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_12.y = RectifiedLinear(rn1_12.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.c = Convolution(rn1_13.c1.c.W[16 x 144], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c1.c.c.y = BatchNormalization(rn1_13.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c1.c.c.sc[16 x 1], rn1_13.c1.c.c.b[16 x 1], rn1_13.c1.c.c.m[16 x 1], rn1_13.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c1.y = RectifiedLinear(rn1_13.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.c = Convolution(rn1_13.c2.W[16 x 144], rn1_13.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_13.c2.c.y = BatchNormalization(rn1_13.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.c2.c.sc[16 x 1], rn1_13.c2.c.b[16 x 1], rn1_13.c2.c.m[16 x 1], rn1_13.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.p = Plus(rn1_13.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_12.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_13.y = RectifiedLinear(rn1_13.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.c = Convolution(rn1_14.c1.c.W[16 x 144], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c1.c.c.y = BatchNormalization(rn1_14.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c1.c.c.sc[16 x 1], rn1_14.c1.c.c.b[16 x 1], rn1_14.c1.c.c.m[16 x 1], rn1_14.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c1.y = RectifiedLinear(rn1_14.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.c = Convolution(rn1_14.c2.W[16 x 144], rn1_14.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_14.c2.c.y = BatchNormalization(rn1_14.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.c2.c.sc[16 x 1], rn1_14.c2.c.b[16 x 1], rn1_14.c2.c.m[16 x 1], rn1_14.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.p = Plus(rn1_14.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_13.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_14.y = RectifiedLinear(rn1_14.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.c = Convolution(rn1_15.c1.c.W[16 x 144], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c1.c.c.y = BatchNormalization(rn1_15.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c1.c.c.sc[16 x 1], rn1_15.c1.c.c.b[16 x 1], rn1_15.c1.c.c.m[16 x 1], rn1_15.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c1.y = RectifiedLinear(rn1_15.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.c = Convolution(rn1_15.c2.W[16 x 144], rn1_15.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_15.c2.c.y = BatchNormalization(rn1_15.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.c2.c.sc[16 x 1], rn1_15.c2.c.b[16 x 1], rn1_15.c2.c.m[16 x 1], rn1_15.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.p = Plus(rn1_15.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_14.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_15.y = RectifiedLinear(rn1_15.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.c = Convolution(rn1_16.c1.c.W[16 x 144], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c1.c.c.y = BatchNormalization(rn1_16.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c1.c.c.sc[16 x 1], rn1_16.c1.c.c.b[16 x 1], rn1_16.c1.c.c.m[16 x 1], rn1_16.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c1.y = RectifiedLinear(rn1_16.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.c = Convolution(rn1_16.c2.W[16 x 144], rn1_16.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_16.c2.c.y = BatchNormalization(rn1_16.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.c2.c.sc[16 x 1], rn1_16.c2.c.b[16 x 1], rn1_16.c2.c.m[16 x 1], rn1_16.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.p = Plus(rn1_16.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_15.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_16.y = RectifiedLinear(rn1_16.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.c = Convolution(rn1_17.c1.c.W[16 x 144], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c1.c.c.y = BatchNormalization(rn1_17.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c1.c.c.sc[16 x 1], rn1_17.c1.c.c.b[16 x 1], rn1_17.c1.c.c.m[16 x 1], rn1_17.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c1.y = RectifiedLinear(rn1_17.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.c = Convolution(rn1_17.c2.W[16 x 144], rn1_17.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_17.c2.c.y = BatchNormalization(rn1_17.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.c2.c.sc[16 x 1], rn1_17.c2.c.b[16 x 1], rn1_17.c2.c.m[16 x 1], rn1_17.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.p = Plus(rn1_17.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_16.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_17.y = RectifiedLinear(rn1_17.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.c = Convolution(rn1_18.c1.c.W[16 x 144], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.c.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c1.c.c.y = BatchNormalization(rn1_18.c1.c.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c1.c.c.sc[16 x 1], rn1_18.c1.c.c.b[16 x 1], rn1_18.c1.c.c.m[16 x 1], rn1_18.c1.c.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c1.y = RectifiedLinear(rn1_18.c1.c.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.c = Convolution(rn1_18.c2.W[16 x 144], rn1_18.c1.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.c2.c.sc = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.b = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.m = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.isd = LearnableParameter -> [16 x 1]
+Validating --> rn1_18.c2.c.y = BatchNormalization(rn1_18.c2.c.c[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_18.c2.c.sc[16 x 1], rn1_18.c2.c.b[16 x 1], rn1_18.c2.c.m[16 x 1], rn1_18.c2.c.isd[16 x 1]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.p = Plus(rn1_18.c2.c.y[32 x 32 x 16 x * {W=32, H=16, C=32}], rn1_17.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn1_18.y = RectifiedLinear(rn1_18.p[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [32 x 32 x 16 x *]
+Validating --> rn2_1.c1.c.c.c = Convolution(rn2_1.c1.c.W[32 x 144], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c1.c.c.y = BatchNormalization(rn2_1.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c1.c.c.sc[32 x 1], rn2_1.c1.c.c.b[32 x 1], rn2_1.c1.c.c.m[32 x 1], rn2_1.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c1.y = RectifiedLinear(rn2_1.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.c = Convolution(rn2_1.c2.W[32 x 288], rn2_1.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c2.c.y = BatchNormalization(rn2_1.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c2.c.sc[32 x 1], rn2_1.c2.c.b[32 x 1], rn2_1.c2.c.m[32 x 1], rn2_1.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1_Wproj = LearnableParameter -> [32 x 16]
+Validating --> rn2_1.c_proj.c = Convolution(rn2_1_Wproj[32 x 16], rn1_18.y[32 x 32 x 16 x * {W=32, H=16, C=32}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.c_proj.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_1.c_proj.y = BatchNormalization(rn2_1.c_proj.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.sc[32 x 1], rn2_1.c_proj.b[32 x 1], rn2_1.c_proj.m[32 x 1], rn2_1.c_proj.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.p = Plus(rn2_1.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.c_proj.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_1.y = RectifiedLinear(rn2_1.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.c = Convolution(rn2_2.c1.c.W[32 x 288], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c1.c.c.y = BatchNormalization(rn2_2.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c1.c.c.sc[32 x 1], rn2_2.c1.c.c.b[32 x 1], rn2_2.c1.c.c.m[32 x 1], rn2_2.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c1.y = RectifiedLinear(rn2_2.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.c = Convolution(rn2_2.c2.W[32 x 288], rn2_2.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_2.c2.c.y = BatchNormalization(rn2_2.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.c2.c.sc[32 x 1], rn2_2.c2.c.b[32 x 1], rn2_2.c2.c.m[32 x 1], rn2_2.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.p = Plus(rn2_2.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_2.y = RectifiedLinear(rn2_2.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.c = Convolution(rn2_3.c1.c.W[32 x 288], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c1.c.c.y = BatchNormalization(rn2_3.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c1.c.c.sc[32 x 1], rn2_3.c1.c.c.b[32 x 1], rn2_3.c1.c.c.m[32 x 1], rn2_3.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c1.y = RectifiedLinear(rn2_3.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.c = Convolution(rn2_3.c2.W[32 x 288], rn2_3.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_3.c2.c.y = BatchNormalization(rn2_3.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.c2.c.sc[32 x 1], rn2_3.c2.c.b[32 x 1], rn2_3.c2.c.m[32 x 1], rn2_3.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.p = Plus(rn2_3.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_2.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_3.y = RectifiedLinear(rn2_3.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.c = Convolution(rn2_4.c1.c.W[32 x 288], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c1.c.c.y = BatchNormalization(rn2_4.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c1.c.c.sc[32 x 1], rn2_4.c1.c.c.b[32 x 1], rn2_4.c1.c.c.m[32 x 1], rn2_4.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c1.y = RectifiedLinear(rn2_4.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.c = Convolution(rn2_4.c2.W[32 x 288], rn2_4.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_4.c2.c.y = BatchNormalization(rn2_4.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.c2.c.sc[32 x 1], rn2_4.c2.c.b[32 x 1], rn2_4.c2.c.m[32 x 1], rn2_4.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.p = Plus(rn2_4.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_3.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_4.y = RectifiedLinear(rn2_4.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.c = Convolution(rn2_5.c1.c.W[32 x 288], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c1.c.c.y = BatchNormalization(rn2_5.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c1.c.c.sc[32 x 1], rn2_5.c1.c.c.b[32 x 1], rn2_5.c1.c.c.m[32 x 1], rn2_5.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c1.y = RectifiedLinear(rn2_5.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.c = Convolution(rn2_5.c2.W[32 x 288], rn2_5.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_5.c2.c.y = BatchNormalization(rn2_5.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.c2.c.sc[32 x 1], rn2_5.c2.c.b[32 x 1], rn2_5.c2.c.m[32 x 1], rn2_5.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.p = Plus(rn2_5.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_4.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_5.y = RectifiedLinear(rn2_5.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.c = Convolution(rn2_6.c1.c.W[32 x 288], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c1.c.c.y = BatchNormalization(rn2_6.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c1.c.c.sc[32 x 1], rn2_6.c1.c.c.b[32 x 1], rn2_6.c1.c.c.m[32 x 1], rn2_6.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c1.y = RectifiedLinear(rn2_6.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.c = Convolution(rn2_6.c2.W[32 x 288], rn2_6.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_6.c2.c.y = BatchNormalization(rn2_6.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.c2.c.sc[32 x 1], rn2_6.c2.c.b[32 x 1], rn2_6.c2.c.m[32 x 1], rn2_6.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.p = Plus(rn2_6.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_5.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_6.y = RectifiedLinear(rn2_6.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.c = Convolution(rn2_7.c1.c.W[32 x 288], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c1.c.c.y = BatchNormalization(rn2_7.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c1.c.c.sc[32 x 1], rn2_7.c1.c.c.b[32 x 1], rn2_7.c1.c.c.m[32 x 1], rn2_7.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c1.y = RectifiedLinear(rn2_7.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.c = Convolution(rn2_7.c2.W[32 x 288], rn2_7.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_7.c2.c.y = BatchNormalization(rn2_7.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.c2.c.sc[32 x 1], rn2_7.c2.c.b[32 x 1], rn2_7.c2.c.m[32 x 1], rn2_7.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.p = Plus(rn2_7.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_6.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_7.y = RectifiedLinear(rn2_7.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.c = Convolution(rn2_8.c1.c.W[32 x 288], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c1.c.c.y = BatchNormalization(rn2_8.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c1.c.c.sc[32 x 1], rn2_8.c1.c.c.b[32 x 1], rn2_8.c1.c.c.m[32 x 1], rn2_8.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c1.y = RectifiedLinear(rn2_8.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.c = Convolution(rn2_8.c2.W[32 x 288], rn2_8.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_8.c2.c.y = BatchNormalization(rn2_8.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.c2.c.sc[32 x 1], rn2_8.c2.c.b[32 x 1], rn2_8.c2.c.m[32 x 1], rn2_8.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.p = Plus(rn2_8.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_7.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_8.y = RectifiedLinear(rn2_8.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.c = Convolution(rn2_9.c1.c.W[32 x 288], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c1.c.c.y = BatchNormalization(rn2_9.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c1.c.c.sc[32 x 1], rn2_9.c1.c.c.b[32 x 1], rn2_9.c1.c.c.m[32 x 1], rn2_9.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c1.y = RectifiedLinear(rn2_9.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.c = Convolution(rn2_9.c2.W[32 x 288], rn2_9.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_9.c2.c.y = BatchNormalization(rn2_9.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.c2.c.sc[32 x 1], rn2_9.c2.c.b[32 x 1], rn2_9.c2.c.m[32 x 1], rn2_9.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.p = Plus(rn2_9.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_8.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_9.y = RectifiedLinear(rn2_9.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.c = Convolution(rn2_10.c1.c.W[32 x 288], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c1.c.c.y = BatchNormalization(rn2_10.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c1.c.c.sc[32 x 1], rn2_10.c1.c.c.b[32 x 1], rn2_10.c1.c.c.m[32 x 1], rn2_10.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c1.y = RectifiedLinear(rn2_10.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.c = Convolution(rn2_10.c2.W[32 x 288], rn2_10.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_10.c2.c.y = BatchNormalization(rn2_10.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.c2.c.sc[32 x 1], rn2_10.c2.c.b[32 x 1], rn2_10.c2.c.m[32 x 1], rn2_10.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.p = Plus(rn2_10.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_9.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_10.y = RectifiedLinear(rn2_10.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.c = Convolution(rn2_11.c1.c.W[32 x 288], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c1.c.c.y = BatchNormalization(rn2_11.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c1.c.c.sc[32 x 1], rn2_11.c1.c.c.b[32 x 1], rn2_11.c1.c.c.m[32 x 1], rn2_11.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c1.y = RectifiedLinear(rn2_11.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.c = Convolution(rn2_11.c2.W[32 x 288], rn2_11.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_11.c2.c.y = BatchNormalization(rn2_11.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.c2.c.sc[32 x 1], rn2_11.c2.c.b[32 x 1], rn2_11.c2.c.m[32 x 1], rn2_11.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.p = Plus(rn2_11.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_10.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_11.y = RectifiedLinear(rn2_11.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.c = Convolution(rn2_12.c1.c.W[32 x 288], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c1.c.c.y = BatchNormalization(rn2_12.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c1.c.c.sc[32 x 1], rn2_12.c1.c.c.b[32 x 1], rn2_12.c1.c.c.m[32 x 1], rn2_12.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c1.y = RectifiedLinear(rn2_12.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.c = Convolution(rn2_12.c2.W[32 x 288], rn2_12.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_12.c2.c.y = BatchNormalization(rn2_12.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.c2.c.sc[32 x 1], rn2_12.c2.c.b[32 x 1], rn2_12.c2.c.m[32 x 1], rn2_12.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.p = Plus(rn2_12.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_11.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_12.y = RectifiedLinear(rn2_12.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.c = Convolution(rn2_13.c1.c.W[32 x 288], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c1.c.c.y = BatchNormalization(rn2_13.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c1.c.c.sc[32 x 1], rn2_13.c1.c.c.b[32 x 1], rn2_13.c1.c.c.m[32 x 1], rn2_13.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c1.y = RectifiedLinear(rn2_13.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.c = Convolution(rn2_13.c2.W[32 x 288], rn2_13.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_13.c2.c.y = BatchNormalization(rn2_13.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.c2.c.sc[32 x 1], rn2_13.c2.c.b[32 x 1], rn2_13.c2.c.m[32 x 1], rn2_13.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.p = Plus(rn2_13.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_12.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_13.y = RectifiedLinear(rn2_13.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.c = Convolution(rn2_14.c1.c.W[32 x 288], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c1.c.c.y = BatchNormalization(rn2_14.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c1.c.c.sc[32 x 1], rn2_14.c1.c.c.b[32 x 1], rn2_14.c1.c.c.m[32 x 1], rn2_14.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c1.y = RectifiedLinear(rn2_14.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.c = Convolution(rn2_14.c2.W[32 x 288], rn2_14.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_14.c2.c.y = BatchNormalization(rn2_14.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.c2.c.sc[32 x 1], rn2_14.c2.c.b[32 x 1], rn2_14.c2.c.m[32 x 1], rn2_14.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.p = Plus(rn2_14.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_13.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_14.y = RectifiedLinear(rn2_14.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.c = Convolution(rn2_15.c1.c.W[32 x 288], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c1.c.c.y = BatchNormalization(rn2_15.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c1.c.c.sc[32 x 1], rn2_15.c1.c.c.b[32 x 1], rn2_15.c1.c.c.m[32 x 1], rn2_15.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c1.y = RectifiedLinear(rn2_15.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.c = Convolution(rn2_15.c2.W[32 x 288], rn2_15.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_15.c2.c.y = BatchNormalization(rn2_15.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.c2.c.sc[32 x 1], rn2_15.c2.c.b[32 x 1], rn2_15.c2.c.m[32 x 1], rn2_15.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.p = Plus(rn2_15.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_14.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_15.y = RectifiedLinear(rn2_15.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.c = Convolution(rn2_16.c1.c.W[32 x 288], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c1.c.c.y = BatchNormalization(rn2_16.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c1.c.c.sc[32 x 1], rn2_16.c1.c.c.b[32 x 1], rn2_16.c1.c.c.m[32 x 1], rn2_16.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c1.y = RectifiedLinear(rn2_16.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.c = Convolution(rn2_16.c2.W[32 x 288], rn2_16.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_16.c2.c.y = BatchNormalization(rn2_16.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.c2.c.sc[32 x 1], rn2_16.c2.c.b[32 x 1], rn2_16.c2.c.m[32 x 1], rn2_16.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.p = Plus(rn2_16.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_15.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_16.y = RectifiedLinear(rn2_16.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.c = Convolution(rn2_17.c1.c.W[32 x 288], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c1.c.c.y = BatchNormalization(rn2_17.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c1.c.c.sc[32 x 1], rn2_17.c1.c.c.b[32 x 1], rn2_17.c1.c.c.m[32 x 1], rn2_17.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c1.y = RectifiedLinear(rn2_17.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.c = Convolution(rn2_17.c2.W[32 x 288], rn2_17.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_17.c2.c.y = BatchNormalization(rn2_17.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.c2.c.sc[32 x 1], rn2_17.c2.c.b[32 x 1], rn2_17.c2.c.m[32 x 1], rn2_17.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.p = Plus(rn2_17.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_16.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_17.y = RectifiedLinear(rn2_17.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.c = Convolution(rn2_18.c1.c.W[32 x 288], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.c.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c1.c.c.y = BatchNormalization(rn2_18.c1.c.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c1.c.c.sc[32 x 1], rn2_18.c1.c.c.b[32 x 1], rn2_18.c1.c.c.m[32 x 1], rn2_18.c1.c.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c1.y = RectifiedLinear(rn2_18.c1.c.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.c = Convolution(rn2_18.c2.W[32 x 288], rn2_18.c1.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.c2.c.sc = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.b = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.m = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.isd = LearnableParameter -> [32 x 1]
+Validating --> rn2_18.c2.c.y = BatchNormalization(rn2_18.c2.c.c[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_18.c2.c.sc[32 x 1], rn2_18.c2.c.b[32 x 1], rn2_18.c2.c.m[32 x 1], rn2_18.c2.c.isd[32 x 1]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.p = Plus(rn2_18.c2.c.y[16 x 16 x 32 x * {W=16, H=32, C=16}], rn2_17.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn2_18.y = RectifiedLinear(rn2_18.p[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [16 x 16 x 32 x *]
+Validating --> rn3_1.c1.c.c.c = Convolution(rn3_1.c1.c.W[64 x 288], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c1.c.c.y = BatchNormalization(rn3_1.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c1.c.c.sc[64 x 1], rn3_1.c1.c.c.b[64 x 1], rn3_1.c1.c.c.m[64 x 1], rn3_1.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c1.y = RectifiedLinear(rn3_1.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.c = Convolution(rn3_1.c2.W[64 x 576], rn3_1.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c2.c.y = BatchNormalization(rn3_1.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c2.c.sc[64 x 1], rn3_1.c2.c.b[64 x 1], rn3_1.c2.c.m[64 x 1], rn3_1.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1_Wproj = LearnableParameter -> [64 x 32]
+Validating --> rn3_1.c_proj.c = Convolution(rn3_1_Wproj[64 x 32], rn2_18.y[16 x 16 x 32 x * {W=16, H=32, C=16}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.c_proj.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_1.c_proj.y = BatchNormalization(rn3_1.c_proj.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.sc[64 x 1], rn3_1.c_proj.b[64 x 1], rn3_1.c_proj.m[64 x 1], rn3_1.c_proj.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.p = Plus(rn3_1.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.c_proj.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_1.y = RectifiedLinear(rn3_1.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.c = Convolution(rn3_2.c1.c.W[64 x 576], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c1.c.c.y = BatchNormalization(rn3_2.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c1.c.c.sc[64 x 1], rn3_2.c1.c.c.b[64 x 1], rn3_2.c1.c.c.m[64 x 1], rn3_2.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c1.y = RectifiedLinear(rn3_2.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.c = Convolution(rn3_2.c2.W[64 x 576], rn3_2.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_2.c2.c.y = BatchNormalization(rn3_2.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.c2.c.sc[64 x 1], rn3_2.c2.c.b[64 x 1], rn3_2.c2.c.m[64 x 1], rn3_2.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.p = Plus(rn3_2.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_2.y = RectifiedLinear(rn3_2.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.c = Convolution(rn3_3.c1.c.W[64 x 576], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c1.c.c.y = BatchNormalization(rn3_3.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c1.c.c.sc[64 x 1], rn3_3.c1.c.c.b[64 x 1], rn3_3.c1.c.c.m[64 x 1], rn3_3.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c1.y = RectifiedLinear(rn3_3.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.c = Convolution(rn3_3.c2.W[64 x 576], rn3_3.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_3.c2.c.y = BatchNormalization(rn3_3.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.c2.c.sc[64 x 1], rn3_3.c2.c.b[64 x 1], rn3_3.c2.c.m[64 x 1], rn3_3.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.p = Plus(rn3_3.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_2.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_3.y = RectifiedLinear(rn3_3.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.c = Convolution(rn3_4.c1.c.W[64 x 576], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c1.c.c.y = BatchNormalization(rn3_4.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c1.c.c.sc[64 x 1], rn3_4.c1.c.c.b[64 x 1], rn3_4.c1.c.c.m[64 x 1], rn3_4.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c1.y = RectifiedLinear(rn3_4.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.c = Convolution(rn3_4.c2.W[64 x 576], rn3_4.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_4.c2.c.y = BatchNormalization(rn3_4.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.c2.c.sc[64 x 1], rn3_4.c2.c.b[64 x 1], rn3_4.c2.c.m[64 x 1], rn3_4.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.p = Plus(rn3_4.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_3.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_4.y = RectifiedLinear(rn3_4.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.c = Convolution(rn3_5.c1.c.W[64 x 576], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c1.c.c.y = BatchNormalization(rn3_5.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c1.c.c.sc[64 x 1], rn3_5.c1.c.c.b[64 x 1], rn3_5.c1.c.c.m[64 x 1], rn3_5.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c1.y = RectifiedLinear(rn3_5.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.c = Convolution(rn3_5.c2.W[64 x 576], rn3_5.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_5.c2.c.y = BatchNormalization(rn3_5.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.c2.c.sc[64 x 1], rn3_5.c2.c.b[64 x 1], rn3_5.c2.c.m[64 x 1], rn3_5.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.p = Plus(rn3_5.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_4.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_5.y = RectifiedLinear(rn3_5.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.c = Convolution(rn3_6.c1.c.W[64 x 576], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c1.c.c.y = BatchNormalization(rn3_6.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c1.c.c.sc[64 x 1], rn3_6.c1.c.c.b[64 x 1], rn3_6.c1.c.c.m[64 x 1], rn3_6.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c1.y = RectifiedLinear(rn3_6.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.c = Convolution(rn3_6.c2.W[64 x 576], rn3_6.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_6.c2.c.y = BatchNormalization(rn3_6.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.c2.c.sc[64 x 1], rn3_6.c2.c.b[64 x 1], rn3_6.c2.c.m[64 x 1], rn3_6.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.p = Plus(rn3_6.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_5.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_6.y = RectifiedLinear(rn3_6.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.c = Convolution(rn3_7.c1.c.W[64 x 576], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c1.c.c.y = BatchNormalization(rn3_7.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c1.c.c.sc[64 x 1], rn3_7.c1.c.c.b[64 x 1], rn3_7.c1.c.c.m[64 x 1], rn3_7.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c1.y = RectifiedLinear(rn3_7.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.c = Convolution(rn3_7.c2.W[64 x 576], rn3_7.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_7.c2.c.y = BatchNormalization(rn3_7.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.c2.c.sc[64 x 1], rn3_7.c2.c.b[64 x 1], rn3_7.c2.c.m[64 x 1], rn3_7.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.p = Plus(rn3_7.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_6.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_7.y = RectifiedLinear(rn3_7.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.c = Convolution(rn3_8.c1.c.W[64 x 576], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c1.c.c.y = BatchNormalization(rn3_8.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c1.c.c.sc[64 x 1], rn3_8.c1.c.c.b[64 x 1], rn3_8.c1.c.c.m[64 x 1], rn3_8.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c1.y = RectifiedLinear(rn3_8.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.c = Convolution(rn3_8.c2.W[64 x 576], rn3_8.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_8.c2.c.y = BatchNormalization(rn3_8.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.c2.c.sc[64 x 1], rn3_8.c2.c.b[64 x 1], rn3_8.c2.c.m[64 x 1], rn3_8.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.p = Plus(rn3_8.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_7.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_8.y = RectifiedLinear(rn3_8.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.c = Convolution(rn3_9.c1.c.W[64 x 576], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c1.c.c.y = BatchNormalization(rn3_9.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c1.c.c.sc[64 x 1], rn3_9.c1.c.c.b[64 x 1], rn3_9.c1.c.c.m[64 x 1], rn3_9.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c1.y = RectifiedLinear(rn3_9.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.c = Convolution(rn3_9.c2.W[64 x 576], rn3_9.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_9.c2.c.y = BatchNormalization(rn3_9.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.c2.c.sc[64 x 1], rn3_9.c2.c.b[64 x 1], rn3_9.c2.c.m[64 x 1], rn3_9.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.p = Plus(rn3_9.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_8.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_9.y = RectifiedLinear(rn3_9.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.c = Convolution(rn3_10.c1.c.W[64 x 576], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c1.c.c.y = BatchNormalization(rn3_10.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c1.c.c.sc[64 x 1], rn3_10.c1.c.c.b[64 x 1], rn3_10.c1.c.c.m[64 x 1], rn3_10.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c1.y = RectifiedLinear(rn3_10.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.c = Convolution(rn3_10.c2.W[64 x 576], rn3_10.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_10.c2.c.y = BatchNormalization(rn3_10.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.c2.c.sc[64 x 1], rn3_10.c2.c.b[64 x 1], rn3_10.c2.c.m[64 x 1], rn3_10.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.p = Plus(rn3_10.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_9.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_10.y = RectifiedLinear(rn3_10.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.c = Convolution(rn3_11.c1.c.W[64 x 576], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c1.c.c.y = BatchNormalization(rn3_11.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c1.c.c.sc[64 x 1], rn3_11.c1.c.c.b[64 x 1], rn3_11.c1.c.c.m[64 x 1], rn3_11.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c1.y = RectifiedLinear(rn3_11.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.c = Convolution(rn3_11.c2.W[64 x 576], rn3_11.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_11.c2.c.y = BatchNormalization(rn3_11.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.c2.c.sc[64 x 1], rn3_11.c2.c.b[64 x 1], rn3_11.c2.c.m[64 x 1], rn3_11.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.p = Plus(rn3_11.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_10.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_11.y = RectifiedLinear(rn3_11.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.c = Convolution(rn3_12.c1.c.W[64 x 576], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c1.c.c.y = BatchNormalization(rn3_12.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c1.c.c.sc[64 x 1], rn3_12.c1.c.c.b[64 x 1], rn3_12.c1.c.c.m[64 x 1], rn3_12.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c1.y = RectifiedLinear(rn3_12.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.c = Convolution(rn3_12.c2.W[64 x 576], rn3_12.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_12.c2.c.y = BatchNormalization(rn3_12.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.c2.c.sc[64 x 1], rn3_12.c2.c.b[64 x 1], rn3_12.c2.c.m[64 x 1], rn3_12.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.p = Plus(rn3_12.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_11.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_12.y = RectifiedLinear(rn3_12.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.c = Convolution(rn3_13.c1.c.W[64 x 576], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c1.c.c.y = BatchNormalization(rn3_13.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c1.c.c.sc[64 x 1], rn3_13.c1.c.c.b[64 x 1], rn3_13.c1.c.c.m[64 x 1], rn3_13.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c1.y = RectifiedLinear(rn3_13.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.c = Convolution(rn3_13.c2.W[64 x 576], rn3_13.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_13.c2.c.y = BatchNormalization(rn3_13.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.c2.c.sc[64 x 1], rn3_13.c2.c.b[64 x 1], rn3_13.c2.c.m[64 x 1], rn3_13.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.p = Plus(rn3_13.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_12.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_13.y = RectifiedLinear(rn3_13.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.c = Convolution(rn3_14.c1.c.W[64 x 576], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c1.c.c.y = BatchNormalization(rn3_14.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c1.c.c.sc[64 x 1], rn3_14.c1.c.c.b[64 x 1], rn3_14.c1.c.c.m[64 x 1], rn3_14.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c1.y = RectifiedLinear(rn3_14.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.c = Convolution(rn3_14.c2.W[64 x 576], rn3_14.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_14.c2.c.y = BatchNormalization(rn3_14.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.c2.c.sc[64 x 1], rn3_14.c2.c.b[64 x 1], rn3_14.c2.c.m[64 x 1], rn3_14.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.p = Plus(rn3_14.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_13.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_14.y = RectifiedLinear(rn3_14.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.c = Convolution(rn3_15.c1.c.W[64 x 576], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c1.c.c.y = BatchNormalization(rn3_15.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c1.c.c.sc[64 x 1], rn3_15.c1.c.c.b[64 x 1], rn3_15.c1.c.c.m[64 x 1], rn3_15.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c1.y = RectifiedLinear(rn3_15.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.c = Convolution(rn3_15.c2.W[64 x 576], rn3_15.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_15.c2.c.y = BatchNormalization(rn3_15.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.c2.c.sc[64 x 1], rn3_15.c2.c.b[64 x 1], rn3_15.c2.c.m[64 x 1], rn3_15.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.p = Plus(rn3_15.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_14.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_15.y = RectifiedLinear(rn3_15.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.c = Convolution(rn3_16.c1.c.W[64 x 576], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c1.c.c.y = BatchNormalization(rn3_16.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c1.c.c.sc[64 x 1], rn3_16.c1.c.c.b[64 x 1], rn3_16.c1.c.c.m[64 x 1], rn3_16.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c1.y = RectifiedLinear(rn3_16.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.c = Convolution(rn3_16.c2.W[64 x 576], rn3_16.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_16.c2.c.y = BatchNormalization(rn3_16.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.c2.c.sc[64 x 1], rn3_16.c2.c.b[64 x 1], rn3_16.c2.c.m[64 x 1], rn3_16.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.p = Plus(rn3_16.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_15.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_16.y = RectifiedLinear(rn3_16.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.c = Convolution(rn3_17.c1.c.W[64 x 576], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c1.c.c.y = BatchNormalization(rn3_17.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c1.c.c.sc[64 x 1], rn3_17.c1.c.c.b[64 x 1], rn3_17.c1.c.c.m[64 x 1], rn3_17.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c1.y = RectifiedLinear(rn3_17.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.c = Convolution(rn3_17.c2.W[64 x 576], rn3_17.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_17.c2.c.y = BatchNormalization(rn3_17.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.c2.c.sc[64 x 1], rn3_17.c2.c.b[64 x 1], rn3_17.c2.c.m[64 x 1], rn3_17.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.p = Plus(rn3_17.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_16.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_17.y = RectifiedLinear(rn3_17.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.c = Convolution(rn3_18.c1.c.W[64 x 576], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.c.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c1.c.c.y = BatchNormalization(rn3_18.c1.c.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c1.c.c.sc[64 x 1], rn3_18.c1.c.c.b[64 x 1], rn3_18.c1.c.c.m[64 x 1], rn3_18.c1.c.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c1.y = RectifiedLinear(rn3_18.c1.c.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.c = Convolution(rn3_18.c2.W[64 x 576], rn3_18.c1.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.c2.c.sc = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.b = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.m = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.isd = LearnableParameter -> [64 x 1]
+Validating --> rn3_18.c2.c.y = BatchNormalization(rn3_18.c2.c.c[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_18.c2.c.sc[64 x 1], rn3_18.c2.c.b[64 x 1], rn3_18.c2.c.m[64 x 1], rn3_18.c2.c.isd[64 x 1]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.p = Plus(rn3_18.c2.c.y[8 x 8 x 64 x * {W=8, H=64, C=8}], rn3_17.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> rn3_18.y = RectifiedLinear(rn3_18.p[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [8 x 8 x 64 x *]
+Validating --> pool = AveragePooling(rn3_18.y[8 x 8 x 64 x * {W=8, H=64, C=8}]) -> [1 x 1 x 64 x *]
+Validating --> OutputNodes.t = Times(OutputNodes.W[10 x 64], pool[1 x 1 x 64 x *]) -> [10 x *]
+Validating --> OutputNodes.b = LearnableParameter -> [10]
+Validating --> OutputNodes.z = Plus(OutputNodes.t[10 x *], OutputNodes.b[10]) -> [10 x *]
+Validating --> CE = CrossEntropyWithSoftmax(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+Validating --> Err = ErrorPrediction(labels[10 x *], OutputNodes.z[10 x *]) -> [1]
+
+559 out of 949 nodes do not share the minibatch layout with the input data.
+
+Post-processing network complete.
+evalNodeNames are not specified, using all the default evalnodes and training criterion nodes.
+
+
+Allocating matrices for forward and/or backward propagation.
+Minibatch[1-20]: Samples Seen = 10000    Err: ErrorPrediction/Sample = 0.0644    CE: CrossEntropyWithSoftmax/Sample = 0.3034767    
+Final Results: Minibatch[1-20]: Samples Seen = 10000    Err: ErrorPrediction/Sample = 0.0644    CE: CrossEntropyWithSoftmax/Sample = 0.3034767    Perplexity = 1.35456    
+__COMPLETED__
diff --git a/Source/CNTK/CNTK.cpp b/Source/CNTK/CNTK.cpp
index 1e0af6dee..f16d26d16 100644
--- a/Source/CNTK/CNTK.cpp
+++ b/Source/CNTK/CNTK.cpp
@@ -70,7 +70,7 @@ void TestCn(const ConfigParameters& config);
 
 void RedirectStdErr(wstring logpath)
 {
-    fprintf(stderr, "Redirecting stderr to file %S\n", logpath.c_str());
+    LOGPRINTF(stderr, "Redirecting stderr to file %S\n", logpath.c_str());
     auto f = make_shared<File>(logpath.c_str(), fileOptionsWrite | fileOptionsText);
     if (dup2(fileno(*f), 2) == -1)
     {
@@ -165,7 +165,7 @@ void DoCommands(const ConfigParameters& config, const shared_ptr<MPIWrapper>& mp
 
     if (numCPUThreads > 0)
     {
-        std::cerr << "Using " << numCPUThreads << " CPU threads." << endl;
+        LOGPRINTF(stderr, "Using %d CPU threads.\n", numCPUThreads);
     }
 
     bool progressTracing = config(L"progressTracing", false);
@@ -187,14 +187,14 @@ void DoCommands(const ConfigParameters& config, const shared_ptr<MPIWrapper>& mp
             if (action[j] == "train" || action[j] == "trainRNN")
             {
                 wstring modelPath = commandParams("modelPath");
-                std::wcerr << "CNTKModelPath: " << modelPath << endl;
+                LOGPRINTF(stderr, "CNTKModelPath: %ls\n", modelPath.c_str());
                 size_t maxEpochs = GetMaxEpochs(commandParams);
-                std::cerr << "CNTKCommandTrainInfo: " + command[i] << " : " << maxEpochs << endl;
+                LOGPRINTF(stderr, "CNTKCommandTrainInfo: %s : %d\n", command[i].c_str(), (int) maxEpochs);
                 fullTotalMaxEpochs += maxEpochs;
             }
         }
     }
-    std::cerr << "CNTKCommandTrainInfo: CNTKNoMoreCommands_Total : " << fullTotalMaxEpochs << endl;
+    LOGPRINTF(stderr, "CNTKCommandTrainInfo: CNTKNoMoreCommands_Total : %d\n", (int) fullTotalMaxEpochs);
 
     // set up progress tracing for compute cluster management
     if (progressTracing && (!mpi || mpi->IsMainNode()))
@@ -225,19 +225,20 @@ void DoCommands(const ConfigParameters& config, const shared_ptr<MPIWrapper>& mp
             // print a banner to visually separate each action in the log
             const char* delim = "##############################################################################";
             const char* prefix = "Action ";
-            fprintf(stderr, "\n%s\n", delim);
-            fprintf(stderr, "#%*s#\n", (int)(strlen(delim) - 2), "");
-            fprintf(stderr, "# %s\"%s\"%*s #\n", prefix, thisAction.c_str(), (int)(strlen(delim) - strlen(prefix) - thisAction.size() - 6), "");
-            fprintf(stderr, "#%*s#\n", (int)(strlen(delim) - 2), "");
-            fprintf(stderr, "%s\n\n", delim);
+            fprintf(stderr, "\n");
+            LOGPRINTF(stderr, "%s\n", delim);
+            LOGPRINTF(stderr, "#%*s#\n", (int)(strlen(delim) - 2), "");
+            LOGPRINTF(stderr, "# %s\"%s\"%*s #\n", prefix, thisAction.c_str(), (int)(strlen(delim) - strlen(prefix) - thisAction.size() - 6), "");
+            LOGPRINTF(stderr, "#%*s#\n", (int)(strlen(delim) - 2), "");
+            LOGPRINTF(stderr, "%s\n\n", delim);
 
             if ((mpi == nullptr) || (commandstoRunOnAllRanks.find(thisAction) != commandstoRunOnAllRanks.end()) || mpi->IsMainNode())
             {
                 if (thisAction == "train" || thisAction == "trainRNN")
                 {
-                    std::cerr << "CNTKCommandTrainBegin: " + command[i] << endl;
+                    LOGPRINTF(stderr, "CNTKCommandTrainBegin: %s\n", command[i].c_str());
                     DoTrain<ConfigParameters, ElemType>(commandParams);
-                    std::cerr << "CNTKCommandTrainEnd: " + command[i] << endl;
+                    LOGPRINTF(stderr, "CNTKCommandTrainEnd: %s\n", command[i].c_str());
                     fullEpochsOffset += GetMaxEpochs(commandParams);
                 }
                 else if (thisAction == "adapt")
@@ -298,7 +299,8 @@ void DoCommands(const ConfigParameters& config, const shared_ptr<MPIWrapper>& mp
                 }
             }
 
-            fprintf(stderr, "\nAction \"%s\" complete.\n\n", thisAction.c_str());
+            fprintf(stderr, "\n");
+            LOGPRINTF(stderr, "Action \"%s\" complete.\n\n", thisAction.c_str());
 
             NDLScript<ElemType> ndlScript;
             ndlScript.ClearGlobal(); // clear global macros between commands
@@ -321,51 +323,51 @@ std::string TimeDateStamp()
 
 void PrintBuiltInfo()
 {
-    fprintf(stderr, "-------------------------------------------------------------------\n");
-    fprintf(stderr, "Build info: \n\n");
-    fprintf(stderr, "\t\tBuilt time: %s %s\n", __DATE__, __TIME__);
-    fprintf(stderr, "\t\tLast modified date: %s\n", __TIMESTAMP__);
+    LOGPRINTF(stderr, "-------------------------------------------------------------------\n");
+    LOGPRINTF(stderr, "Build info: \n\n");
+    LOGPRINTF(stderr, "\t\tBuilt time: %s %s\n", __DATE__, __TIME__);
+    LOGPRINTF(stderr, "\t\tLast modified date: %s\n", __TIMESTAMP__);
 #ifdef _BUILDTYPE_
-    fprintf(stderr, "\t\tBuild type: %s\n", _BUILDTYPE_);
+    LOGPRINTF(stderr, "\t\tBuild type: %s\n", _BUILDTYPE_);
 #endif
 #ifdef _BUILDTARGET_
-    fprintf(stderr, "\t\tBuild target: %s\n", _BUILDTARGET_);
+    LOGPRINTF(stderr, "\t\tBuild target: %s\n", _BUILDTARGET_);
 #endif
 #ifdef _WITH_1BITSGD_
-    fprintf(stderr, "\t\tWith 1bit-SGD: %s\n", _WITH_1BITSGD_);
+    LOGPRINTF(stderr, "\t\tWith 1bit-SGD: %s\n", _WITH_1BITSGD_);
 #endif
 #ifdef _MATHLIB_
-    fprintf(stderr, "\t\tMath lib: %s\n", _MATHLIB_);
+    LOGPRINTF(stderr, "\t\tMath lib: %s\n", _MATHLIB_);
 #endif
 #ifdef _CUDA_PATH_
-    fprintf(stderr, "\t\tCUDA_PATH: %s\n", _CUDA_PATH_);
+    LOGPRINTF(stderr, "\t\tCUDA_PATH: %s\n", _CUDA_PATH_);
 #endif
 #ifdef _CUB_PATH_
-    fprintf(stderr, "\t\tCUB_PATH: %s\n", _CUB_PATH_);
+    LOGPRINTF(stderr, "\t\tCUB_PATH: %s\n", _CUB_PATH_);
 #endif
 #ifdef _CUDNN_PATH_
-    fprintf(stderr, "\t\tCUDNN_PATH: %s\n", _CUDNN_PATH_);
+    LOGPRINTF(stderr, "\t\tCUDNN_PATH: %s\n", _CUDNN_PATH_);
 #endif
 #ifdef _GIT_EXIST
-    fprintf(stderr, "\t\tBuild Branch: %s\n", _BUILDBRANCH_);
-    fprintf(stderr, "\t\tBuild SHA1: %s\n", _BUILDSHA1_);
+    LOGPRINTF(stderr, "\t\tBuild Branch: %s\n", _BUILDBRANCH_);
+    LOGPRINTF(stderr, "\t\tBuild SHA1: %s\n", _BUILDSHA1_);
 #endif
 #ifdef _BUILDER_
-    fprintf(stderr, "\t\tBuilt by %s on %s\n", _BUILDER_, _BUILDMACHINE_);
+    LOGPRINTF(stderr, "\t\tBuilt by %s on %s\n", _BUILDER_, _BUILDMACHINE_);
 #endif
 #ifdef _BUILDPATH_
-    fprintf(stderr, "\t\tBuild Path: %s\n", _BUILDPATH_);
+    LOGPRINTF(stderr, "\t\tBuild Path: %s\n", _BUILDPATH_);
 #endif
-    fprintf(stderr, "-------------------------------------------------------------------\n");
+    LOGPRINTF(stderr, "-------------------------------------------------------------------\n");
 }
 
 void PrintUsageInfo()
 {
-    fprintf(stderr, "-------------------------------------------------------------------\n");
-    fprintf(stderr, "Usage: cntk configFile=yourConfigFile\n");
-    fprintf(stderr, "For detailed information please consult the CNTK book\n");
-    fprintf(stderr, "\"An Introduction to Computational Networks and the Computational Network Toolkit\"\n");
-    fprintf(stderr, "-------------------------------------------------------------------\n");
+    LOGPRINTF(stderr, "-------------------------------------------------------------------\n");
+    LOGPRINTF(stderr, "Usage: cntk configFile=yourConfigFile\n");
+    LOGPRINTF(stderr, "For detailed information please consult the CNTK book\n");
+    LOGPRINTF(stderr, "\"An Introduction to Computational Networks and the Computational Network Toolkit\"\n");
+    LOGPRINTF(stderr, "-------------------------------------------------------------------\n");
 }
 
 // ---------------------------------------------------------------------------
@@ -412,9 +414,11 @@ int wmainWithBS(int argc, wchar_t* argv[]) // called from wmain which is a wrapp
     wstring startupMessage = msra::strfun::wstrprintf(L"running on %ls at %ls\n", msra::strfun::utf16(GetHostName()).c_str(), msra::strfun::utf16(TimeDateStamp()).c_str());
     startupMessage += msra::strfun::wstrprintf(L"command line: %ls", exePath.c_str());
     for (const auto& arg : args)
+    {
         startupMessage += L"  " + arg;
+    }
 
-    fprintf(stderr, "%ls\n", startupMessage.c_str());
+    LOGPRINTF(stderr, "%ls\n", startupMessage.c_str());
 
     // parse command-line options
     vector<wstring> sourceFiles;
@@ -443,6 +447,7 @@ int wmainWithBS(int argc, wchar_t* argv[]) // called from wmain which is a wrapp
     // compile the BrainScript
     wstring bs = L"[\n";
     bs += L"include \'cntk.core.bs'"; // start with including the standard macros
+
     // Note: Using lowercase ^^ here to match the Linux name of the CNTK exe.
     //bs += standardFunctions + computationNodes + commonMacros + L"\n";
     for (const auto& sourceFile : sourceFiles)
@@ -451,7 +456,8 @@ int wmainWithBS(int argc, wchar_t* argv[]) // called from wmain which is a wrapp
     for (const auto& over : overrides)
         bs += L"with [ " + over + L" ]\n";
 
-    fprintf(stderr, "\n\nBrainScript -->\n\n%ls\n\n", bs.c_str());
+    fprintf(stderr, "\n\n");
+    LOGPRINTF(stderr, "BrainScript -->\n\n%ls\n\n", bs.c_str());
 
     let expr = BS::ParseConfigExpression(bs, move(includePaths)); // parse
     let valp = BS::Evaluate(expr);                                // evaluate parse into a dictionary
@@ -486,7 +492,7 @@ int wmainWithBS(int argc, wchar_t* argv[]) // called from wmain which is a wrapp
             logpath += msra::strfun::wstrprintf(L"rank%d", (int) mpi->CurrentNodeRank());
 
         RedirectStdErr(logpath);
-        fprintf(stderr, "%ls\n", startupMessage.c_str());
+        LOGPRINTF(stderr, "%ls\n", startupMessage.c_str());
     }
 
     // echo config info to log
@@ -497,10 +503,11 @@ int wmainWithBS(int argc, wchar_t* argv[]) // called from wmain which is a wrapp
     int numCPUThreads = config(L"numCPUThreads", 0);
     numCPUThreads = CPUMatrix<float /*any will do*/>::SetNumThreads(numCPUThreads);
     if (numCPUThreads > 0)
-        fprintf(stderr, "Using %d CPU threads.\n", numCPUThreads);
+        LOGPRINTF(stderr, "Using %d CPU threads.\n", numCPUThreads);
 
     bool progressTracing = config(L"progressTracing", false);
     size_t fullTotalMaxEpochs = 1; // BUGBUG: BS does not allow me to read out the max epochs parameters, as that would instantiate and thus execute the objects
+
     // set up progress tracing for compute cluster management
     if (progressTracing && ((mpi == nullptr) || mpi->IsMainNode()))
         ProgressTracing::TraceTotalNumberOfSteps(fullTotalMaxEpochs); // enable tracing, using this as the total number of epochs
@@ -532,7 +539,8 @@ int wmainWithBS(int argc, wchar_t* argv[]) // called from wmain which is a wrapp
         fprintf(fp, "successfully finished at %s on %s\n", TimeDateStamp().c_str(), GetHostName().c_str());
         fcloseOrDie(fp);
     }
-    fprintf(stderr, "COMPLETED\n"), fflush(stderr);
+    LOGPRINTF(stderr, "__COMPLETED__\n");
+    fflush(stderr);
 
     MPIWrapper::DeleteInstance();
     return EXIT_SUCCESS;
@@ -585,36 +593,51 @@ int wmainOldCNTKConfig(int argc, wchar_t* argv[]) // called from wmain which is
     PrintBuiltInfo(); // this one goes to log file
     std::string timestamp = TimeDateStamp();
 
+    bool timestamping = config(L"timestamping", false);
+    if (timestamping)
+    {
+        ProgressTracing::SetTimestampingFlag();
+    }
+
     // dump config info
-    fprintf(stderr, "\nRunning on %s at %s\n", GetHostName().c_str(), timestamp.c_str());
-    fprintf(stderr, "Command line: \n");
+    fprintf(stderr, "\n");
+    LOGPRINTF(stderr, "Running on %s at %s\n", GetHostName().c_str(), timestamp.c_str());
+    LOGPRINTF(stderr, "Command line: \n");
     for (int i = 0; i < argc; i++)
-        fprintf(stderr, "%*s%ls", i > 0 ? 2 : 0, "", argv[i]); // use 2 spaces for better visual separability
+    {
+        // use 2 spaces for better visual separability
+        fprintf(stderr, "%*s%ls", i > 0 ? 2 : 0, "", argv[i]);
+    }
     fprintf(stderr, "\n\n");
 
 #if 1 //def _DEBUG
     // This simply merges all the different config parameters specified (eg, via config files or via command line directly),
     // and prints it.
-    fprintf(stderr, "\n\n>>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>\n");
-    fprintf(stderr, "%s\n", rawConfigString.c_str());
-    fprintf(stderr, "<<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<\n");
+    fprintf(stderr, "\n\n");
+    LOGPRINTF(stderr, ">>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>\n");
+    LOGPRINTF(stderr, "%s\n", rawConfigString.c_str());
+    LOGPRINTF(stderr, "<<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<\n");
 
-    // Same as above, but all variables are resolved.  If a parameter is set multiple times (eg, set in config, overriden at command line),
+    // Same as above, but all variables are resolved.  If a parameter is set multiple times (eg, set in config, overridden at command line),
     // All of these assignments will appear, even though only the last assignment matters.
-    fprintf(stderr, "\n>>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>\n");
-    fprintf(stderr, "%s\n", config.ResolveVariables(rawConfigString).c_str());
-    fprintf(stderr, "<<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<\n");
+    fprintf(stderr, "\n");
+    LOGPRINTF(stderr, ">>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>\n");
+    LOGPRINTF(stderr, "%s\n", config.ResolveVariables(rawConfigString).c_str());
+    LOGPRINTF(stderr, "<<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<\n");
 
     // This outputs the final value each variable/parameter is assigned to in config (so if a parameter is set multiple times, only the last
     // value it is set to will appear).
-    fprintf(stderr, "\n>>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>\n");
+    fprintf(stderr, "\n");
+    LOGPRINTF(stderr, ">>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>\n");
     config.dumpWithResolvedVariables();
-    fprintf(stderr, "<<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<\n");
+    LOGPRINTF(stderr, "<<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<\n");
 #endif
 
-    fprintf(stderr, "Commands:");
+    LOGPRINTF(stderr, "Commands:");
     for (int i = 0; i < command.size(); i++)
+    {
         fprintf(stderr, " %s", command[i].c_str());
+    }
     fprintf(stderr, "\n");
 
     // run commands
@@ -623,7 +646,8 @@ int wmainOldCNTKConfig(int argc, wchar_t* argv[]) // called from wmain which is
     if (config.Exists("type"))
         InvalidArgument("CNTK: Use of 'type' parameter is deprecated, it is called 'precision' now.");
 
-    fprintf(stderr, "Precision = \"%s\"\n", type.c_str());
+    LOGPRINTF(stderr, "Precision = \"%s\"\n", type.c_str());
+
     if (type == "float")
         DoCommands<float>(config, mpi);
     else if (type == "double")
@@ -638,7 +662,8 @@ int wmainOldCNTKConfig(int argc, wchar_t* argv[]) // called from wmain which is
         fprintf(fp, "successfully finished at %s on %s\n", TimeDateStamp().c_str(), GetHostName().c_str());
         fcloseOrDie(fp);
     }
-    fprintf(stderr, "COMPLETED\n"), fflush(stderr);
+    LOGPRINTF(stderr, "__COMPLETED__\n");
+    fflush(stderr);
 
     MPIWrapper::DeleteInstance();
     return EXIT_SUCCESS;
@@ -664,38 +689,45 @@ int wmain1(int argc, wchar_t* argv[]) // called from wmain which is a wrapper th
         PrintBuiltInfo(); // print build info directly in case that user provides zero argument (convenient for checking build type)
         if (argc <= 1)
         {
-            fprintf(stderr, "No command-line argument given.\n");
+            LOGPRINTF(stderr, "No command-line argument given.\n");
             PrintUsageInfo();
             return EXIT_FAILURE;
         }
+
         // detect legacy CNTK configuration
         bool isOldCNTKConfig = false;
         for (int i = 0; i < argc && !isOldCNTKConfig; i++)
             isOldCNTKConfig |= !_wcsnicmp(L"configFile=", argv[i], 11);
+
         if (isOldCNTKConfig)
             return wmainOldCNTKConfig(argc, argv);
+
         // run from BrainScript
         return wmainWithBS(argc, argv);
     }
     catch (const ScriptableObjects::ScriptingException& err)
     {
-        fprintf(stderr, "\nEXCEPTION occurred: %s\n", err.what());
+        fprintf(stderr, "\n");
+        LOGPRINTF(stderr, "EXCEPTION occurred: %s\n", err.what());
         err.PrintError();
         return EXIT_FAILURE;
     }
     catch (const IExceptionWithCallStackBase& err)
     {
-        fprintf(stderr, "\nEXCEPTION occurred: %s\n%s", dynamic_cast<const std::exception&>(err).what(), err.CallStack());
+        fprintf(stderr, "\n");
+        LOGPRINTF(stderr, "EXCEPTION occurred: %s\n%s", dynamic_cast<const std::exception&>(err).what(), err.CallStack());
         return EXIT_FAILURE;
     }
     catch (const std::exception& err)
     {
-        fprintf(stderr, "\nEXCEPTION occurred: %s\n", err.what());
+        fprintf(stderr, "\n");
+        LOGPRINTF(stderr, "EXCEPTION occurred: %s\n", err.what());
         return EXIT_FAILURE;
     }
     catch (...)
     {
-        fprintf(stderr, "\nUnknown ERROR occurred\n");
+        fprintf(stderr, "\n");
+        LOGPRINTF(stderr, "Unknown ERROR occurred\n");
         return EXIT_FAILURE;
     }
 }
@@ -703,7 +735,8 @@ int wmain1(int argc, wchar_t* argv[]) // called from wmain which is a wrapper th
 #ifdef __WINDOWS__
 void TerminateThis()
 {
-    fprintf(stderr, "terminate_this: aborting\n"), fflush(stderr);
+    LOGPRINTF(stderr, "terminate_this: aborting\n");
+    fflush(stderr);
     exit(EXIT_FAILURE);
 }
 
@@ -714,7 +747,7 @@ static void LogDelayLoadError(PEXCEPTION_POINTERS pExcPointers)
     if (pExcPointers->ExceptionRecord->ExceptionCode == EXCEPTION_DLL_NOT_FOUND)
     {
         const auto & pDelayLoadInfo = *PDelayLoadInfo(pExcPointers->ExceptionRecord->ExceptionInformation[0]);
-        fprintf(stderr, "CNTK: Failed to load DLL '%s'.\n", pDelayLoadInfo.szDll);
+        LOGPRINTF(stderr, "CNTK: Failed to load DLL '%s'.\n", pDelayLoadInfo.szDll);
     }
 }
 
@@ -736,7 +769,7 @@ int wmain(int argc, wchar_t* argv[]) // wmain wrapper that reports Win32 excepti
         else if (code == EXCEPTION_INT_DIVIDE_BY_ZERO) msg = ": Integer division by zero";
         else if (code == EXCEPTION_STACK_OVERFLOW)     msg = ": Stack overflow";
         else if (code == EXCEPTION_DLL_NOT_FOUND)      msg = ": Module not found";
-        fprintf(stderr, "CNTK: Caught Win32 exception 0x%08x%s.\n", (unsigned int)code, msg);
+        LOGPRINTF(stderr, "CNTK: Caught Win32 exception 0x%08x%s.\n", (unsigned int)code, msg);
         fflush(stderr);
         exit(EXIT_FAILURE);
     }
diff --git a/Source/Common/Include/ProgressTracing.h b/Source/Common/Include/ProgressTracing.h
index fba0b8e3c..f15435431 100644
--- a/Source/Common/Include/ProgressTracing.h
+++ b/Source/Common/Include/ProgressTracing.h
@@ -4,10 +4,32 @@
 //
 #pragma once
 
+#include <chrono>
 #include "TimerUtility.h"
 
 namespace Microsoft { namespace MSR { namespace CNTK {
 
+// If the Tracing flag is set, print out a timestamp with no new line at the end
+#define PREPENDTS(stream) \
+    do \
+    { \
+        if (ProgressTracing::GetTimestampingFlag()) \
+        { \
+           std::time_t tt = std::chrono::system_clock::to_time_t(std::chrono::system_clock::now()); \
+           char mbstr[30]; \
+           if (std::strftime(mbstr, sizeof(mbstr), "%m/%d/%Y %H:%M:%S", std::localtime(&tt))) \
+               fprintf(stream, "%s: ", mbstr);  \
+        } \
+    } while(0)
+
+// Print out a log message.  If the Tracing flag is set, prepend with a timestamp
+#define LOGPRINTF(stream, ...) \
+    do \
+    { \
+        PREPENDTS(stream); \
+        fprintf(stream, __VA_ARGS__); \
+    } while(0)
+
 // ---------------------------------------------------------------------------
 // ProgressTracing -- static helper class for logging a progress indicator
 //
@@ -29,12 +51,13 @@ namespace Microsoft { namespace MSR { namespace CNTK {
 {
     bool m_enabled;
     bool m_tracingFlag;
+    bool m_timestampFlag;
     size_t m_totalNumberOfSteps; // total number of epochs in entire training run
     size_t m_currentStepOffset;  // current offset
     Timer m_progressTracingTimer;
 
     ProgressTracing()
-        : m_enabled(false), m_tracingFlag(false), m_totalNumberOfSteps(0), m_currentStepOffset(0)
+        : m_enabled(false), m_tracingFlag(false), m_timestampFlag(false), m_totalNumberOfSteps(0), m_currentStepOffset(0)
     {
     }
 
@@ -50,12 +73,23 @@ public:
         return GetStaticInstance().m_tracingFlag;
     }
 
+    static bool GetTimestampingFlag()
+    {
+        return GetStaticInstance().m_timestampFlag;
+    }
+
     static void SetTracingFlag()
     {
         auto& us = GetStaticInstance();
         us.m_tracingFlag = true;
     }
 
+    static void SetTimestampingFlag()
+    {
+        auto& us = GetStaticInstance();
+        us.m_timestampFlag = true;
+    }
+
     // call TraceTotalNumberOfSteps() to set the total number of steps
     // Calling this with totalNumberOfSteps>0 will enable progress tracing.
     static void TraceTotalNumberOfSteps(size_t totalNumberOfSteps)
diff --git a/Source/Common/fileutil.cpp b/Source/Common/fileutil.cpp
index a30d89376..d90ae5996 100644
--- a/Source/Common/fileutil.cpp
+++ b/Source/Common/fileutil.cpp
@@ -609,11 +609,6 @@ void renameOrDie(const std::string& from, const std::string& to)
     // WORKAROUND: "rename" should do this but this is a workaround
     // to the HDFS FUSE implementation's bug of failing to do so
     // workaround for FUSE rename when running on Philly
-    if (ProgressTracing::GetTracingFlag())
-    {
-        fprintf(stderr, "rename %s to %s\n", from.c_str(), to.c_str());
-    }    
-    
     unlinkOrDie(to);
     if (rename(from.c_str(), to.c_str()) != 0)
     {
diff --git a/Source/ComputationNetworkLib/ComputationNetworkAnalysis.cpp b/Source/ComputationNetworkLib/ComputationNetworkAnalysis.cpp
index b865f50f0..a60e0655d 100644
--- a/Source/ComputationNetworkLib/ComputationNetworkAnalysis.cpp
+++ b/Source/ComputationNetworkLib/ComputationNetworkAnalysis.cpp
@@ -94,7 +94,9 @@ void ComputationNetwork::FormRecurrentLoops(const ComputationNodeBasePtr& rootNo
                 if (node->Input(i)->m_loopId == node->m_loopId && GetRecurrenceSteppingDirection(node) == 0/*not a Delay node*/)
                 {
                     // assert(node->Input(i)->m_indexInLoop == 0);                    // No. It seems this variable really counts the number of parents.
-                    node->Input(i)->m_indexInLoop++; // BUGBUG: this is bumping up the m_indexInLoop, but I don't think it is initialized anywhere other than PurgeStateForFormingRecurrentLoops(). i-1?
+
+                    // BUGBUG: this is bumping up the m_indexInLoop, but I don't think it is initialized anywhere other than PurgeStateForFormingRecurrentLoops(). i-1?
+                    node->Input(i)->m_indexInLoop++;
                 }
             }
         }
diff --git a/Source/ComputationNetworkLib/ComputationNetworkEvaluation.cpp b/Source/ComputationNetworkLib/ComputationNetworkEvaluation.cpp
index 77e5fac3c..bcbc92b4c 100644
--- a/Source/ComputationNetworkLib/ComputationNetworkEvaluation.cpp
+++ b/Source/ComputationNetworkLib/ComputationNetworkEvaluation.cpp
@@ -114,9 +114,11 @@ ComputationNetwork::PARTraversalFlowControlNode::PARTraversalFlowControlNode(con
         {
             // instead of the node itself, include the sentinel SEQTraversalFlowControlNode in our list
             m_nestedNodes.push_back(recInfo);
+
             // and verify that we only encountered the loop once (all nodes should have been consecutive)
             if (!loopsSeen.insert(recInfo).second)
                 LogicError("PARTraversalFlowControlNode: members of loop %ls are not consecutive in node list.", recInfo->NodeName().c_str());
+
             // consume all nodes that are part of the same loop (they are all consecutive)
             while (nodeIter != allNodes.end() && (*nodeIter)->IsPartOfLoop() && FindInRecurrentLoops(recurrentInfo, *nodeIter) == recInfo)
                 nodeIter++;
@@ -303,8 +305,10 @@ ComputationNetwork::PARTraversalFlowControlNode::PARTraversalFlowControlNode(con
     // look in all recurrent loops of the network
     // TODO: Check for IsPartOfLoop(). Also why not store the loop id in the node for direct lookup?
     for (auto& iter : recurrentInfo)
+    {
         if (std::find(iter->m_nestedNodes.begin(), iter->m_nestedNodes.end(), node) != iter->m_nestedNodes.end()) // TODO: should this loop need to be a method of SEQTraversalFlowControlNode?
             return iter;
+    }
     return nullptr; // not part of a recurrent loop
 }
 
@@ -357,8 +361,10 @@ void ComputationNetwork::PrintComputationTree(const ComputationNodeBasePtr& root
     if (nodes.size() == 0)
         fprintf(stderr, "\n(empty)\n");
     else
+    {
         for (const auto& node : nodes)
             node->PrintSelf(printMatrices);
+    }
 }
 
 // -----------------------------------------------------------------------
@@ -397,9 +403,11 @@ void ComputationNetwork::CompileNetwork()
     // all steps below have to be repeated for all root nodes (=nodes without parents and PreComputeNodes)
     DetermineSetOfAllRoots();
 
-    fprintf(stderr, "\n%d roots:\n", (int) m_allRoots.size());
+    fprintf(stderr, "\n%d roots:\n", (int)m_allRoots.size());
     for (const auto& root : m_allRoots)
+    {
         fprintf(stderr, "\t%ls = %ls\n", root->NodeName().c_str(), root->OperationName().c_str());
+    }
 
     // Note: Steps below are loops over root nodes. We will gradually push those loops through to the functions,
     //       to reduce redundant operation on shared portions of the network.
@@ -473,10 +481,13 @@ void ComputationNetwork::DetermineSetOfAllRoots()
     set<ComputationNodeBasePtr> allKnownRoots;
     for (const auto& node : FinalCriterionNodes())
         allKnownRoots.insert(node);
+
     for (const auto& node : EvaluationNodes())
         allKnownRoots.insert(node);
+
     for (const auto& node : OutputNodes())
         allKnownRoots.insert(node);
+
     for (const auto& iter : m_nameToNodeMap) // PreComputeNodes
     {
         auto node = iter.second;
@@ -513,7 +524,9 @@ void ComputationNetwork::ValidateNetwork()
     // set up MBLayout links of inputs (all others get propagated upwards through Validate())
     // TODO: Once we support mismatching layouts, this will be more involved. For now, everything shares the one layout that the Network knows about.
     for (auto node : InputNodes(nullptr))
+    {
         node->LinkToMBLayout(m_pMBLayout);
+    }
 
     // we call all nodes' Validate() in order to validate, that is, set up MBLayout and FunctionValues dimension
     // A problem is that recurrent loops may require partial validation.
@@ -542,6 +555,7 @@ void ComputationNetwork::ValidateNetwork()
     }
     fprintf(stderr, "\nValidating network, final pass.\n\n");
     toValidate = ValidateNodes(nodes, /*isFirstPass=*/pass == 1, true /*isFinalValidationPass*/);
+
     if (toValidate != 0)
         LogicError("ValidateSubNetwork: ValidateNodes(true) unexpectedly returned with work left to do.");
 
@@ -571,7 +585,7 @@ void ComputationNetwork::ValidateNetwork()
     }
     if (!nonDefaultNodes.empty())
     {
-        fprintf(stderr, "%d out of %d nodes do not share the minibatch layout with the input data.\n", (int) nonDefaultNodes.size(), (int) nodes.size());
+        fprintf(stderr, "%d out of %d nodes do not share the minibatch layout with the input data.\n", (int)nonDefaultNodes.size(), (int)nodes.size());
         // for (auto node : nonDefaultNodes)
         //    fprintf(stderr, "    %ls\n", node->NodeName().c_str());
         // fprintf(stderr, "\n\n");
@@ -631,6 +645,7 @@ size_t ComputationNetwork::ValidateNodes(list<ComputationNodeBasePtr> nodes, boo
             hasVisitedChild |= child->m_visited; // if not a single visited child then no point in validating
             allChildrenVisited &= child->m_visited;
         }
+
         // if there is not at least one visited child
         bool valid = false;
         if (hasVisitedChild || isLeaf) // got at least one child: it makes sense to call Validate()
@@ -652,8 +667,10 @@ size_t ComputationNetwork::ValidateNodes(list<ComputationNodeBasePtr> nodes, boo
             node->m_visited = true;
             // print the new type
             // sanity checks
+
             if (isFinalValidationPass && !unchanged)
                 LogicError("ValidateSubNetwork: %ls %ls operation changed during final validation.", node->NodeName().c_str(), node->OperationName().c_str());
+
             if (isFinalValidationPass && !allChildrenVisited)
                 LogicError("ValidateSubNetwork: %ls %ls operation in final validation although not all children were visited?", node->NodeName().c_str(), node->OperationName().c_str());
             // if all children valid then
@@ -830,7 +847,7 @@ void ComputationNetwork::AllocateAllMatrices(const std::vector<ComputationNodeBa
         else
         {
             nodeIter->RequestMatricesBeforeForwardProp(m_matrixPool);
-            // we only release matrices for the children since the root node's informatioin will be used and should not be shared
+            // we only release matrices for the children since the root node's information will be used and should not be shared
             // with others
             ReleaseMatricesAfterEvalForChildren(nodeIter, parentCount);
         }
diff --git a/Source/SGDLib/SGD.cpp b/Source/SGDLib/SGD.cpp
index 463f2aa37..adfe55d4e 100644
--- a/Source/SGDLib/SGD.cpp
+++ b/Source/SGDLib/SGD.cpp
@@ -44,22 +44,23 @@ void SGD<ElemType>::Train(function<ComputationNetworkPtr(DEVICEID_TYPE)> createN
     int startEpoch = DetermineStartEpoch(makeMode);
     if (startEpoch == m_maxEpochs)
     {
-        fprintf(stderr, "No further training is necessary.\n");
+        LOGPRINTF(stderr, "No further training is necessary.\n");
         return;
     }
 
     wstring modelFileName = GetModelNameForEpoch(int(startEpoch) - 1);
     bool loadNetworkFromCheckpoint = startEpoch >= 0;
+    fprintf(stderr, "\n");
     if (loadNetworkFromCheckpoint)
-        fprintf(stderr, "\nStarting from checkpoint. Loading network from '%ls'.\n", modelFileName.c_str());
+        LOGPRINTF(stderr, "Starting from checkpoint. Loading network from '%ls'.\n", modelFileName.c_str());
     else
-        fprintf(stderr, "\nCreating virgin network.\n");
+        LOGPRINTF(stderr, "Creating virgin network.\n");
 
     // create or load from checkpoint
     shared_ptr<ComputationNetwork> net = !loadNetworkFromCheckpoint ? createNetworkFn(deviceId) : ComputationNetwork::CreateFromFile<ElemType>(deviceId, modelFileName);
 
     // log the device we are computing on
-    fprintf(stderr, "%s model with %d nodes", loadNetworkFromCheckpoint ? "Loaded" : "Created", (int)net->GetTotalNumberOfNodes());
+    LOGPRINTF(stderr, "%s model with %d nodes", loadNetworkFromCheckpoint ? "Loaded" : "Created", (int)net->GetTotalNumberOfNodes());
     if (net->GetDeviceId() < 0)
         fprintf(stderr, " on CPU.\n");
     else
@@ -74,6 +75,7 @@ void SGD<ElemType>::Train(function<ComputationNetworkPtr(DEVICEID_TYPE)> createN
     // set tracing flags
     for (const auto& traceNodeName : m_traceNodeNamesReal)
         net->GetNodeFromName(traceNodeName)->EnableNodeTracing(/*isCategoryLabel=*/false);
+
     for (const auto& traceNodeName : m_traceNodeNamesCategory)
         net->GetNodeFromName(traceNodeName)->EnableNodeTracing(/*isCategoryLabel=*/true);
 
@@ -93,7 +95,7 @@ void SGD<ElemType>::Adapt(wstring origModelFileName, wstring refNodeName,
     int startEpoch = DetermineStartEpoch(makeMode);
     if (startEpoch == m_maxEpochs)
     {
-        fprintf(stderr, "No further training is necessary.\n");
+        LOGPRINTF(stderr, "No further training is necessary.\n");
         return;
     }
 
@@ -102,13 +104,13 @@ void SGD<ElemType>::Adapt(wstring origModelFileName, wstring refNodeName,
     if (startEpoch >= 0)
     {
         wstring modelFileName = GetModelNameForEpoch(int(startEpoch) - 1);
-        fprintf(stderr, "Starting from checkpoint. Loading network from '%ls'.\n", modelFileName.c_str());
+        LOGPRINTF(stderr, "Starting from checkpoint. Loading network from '%ls'.\n", modelFileName.c_str());
         net = ComputationNetwork::CreateFromFile<ElemType>(deviceId, modelFileName);
         networkLoadedFromCheckpoint = true;
     }
     else
     {
-        fprintf(stderr, "Load Network From the original model file %ls.\n", origModelFileName.c_str());
+        LOGPRINTF(stderr, "Load Network From the original model file %ls.\n", origModelFileName.c_str());
         net = ComputationNetwork::CreateFromFile<ElemType>(deviceId, origModelFileName);
     }
 
@@ -118,14 +120,14 @@ void SGD<ElemType>::Adapt(wstring origModelFileName, wstring refNodeName,
     m_needAdaptRegularization = m_adaptationRegType != AdaptationRegType::None && m_adaptationRegWeight > 0;
     if (m_needAdaptRegularization)
     {
-        fprintf(stderr, "Load reference Network From the original model file %ls.\n", origModelFileName.c_str());
+        LOGPRINTF(stderr, "Load reference Network From the original model file %ls.\n", origModelFileName.c_str());
         refNet = ComputationNetwork::CreateFromFile<ElemType>(deviceId, origModelFileName);
     }
 
     ComputationNodeBasePtr refNode;
     if (m_needAdaptRegularization && m_adaptationRegType == AdaptationRegType::KL)
     {
-        fprintf(stderr, "Checking refNodeName %ls.\n", origModelFileName.c_str());
+        LOGPRINTF(stderr, "Checking refNodeName %ls.\n", origModelFileName.c_str());
         if (refNodeName == L"")
             InvalidArgument("refNodeName does not exist and is needed when adaptationRegType is KL.");
         refNode = refNet->GetNodeFromName(refNodeName);
@@ -152,9 +154,12 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
     auto& labelNodes = net->LabelNodes();
     auto& criterionNodes = GetTrainCriterionNodes(net);
 
-    fprintf(stderr, "\nTraining criterion node(s):\n");
+    fprintf(stderr, "\n");
+    LOGPRINTF(stderr, "Training criterion node(s):\n");
     for (const auto& node : criterionNodes)
-        fprintf(stderr, "\t%ls = %ls\n", node->NodeName().c_str(), node->OperationName().c_str());
+    {
+        LOGPRINTF(stderr, "\t%ls = %ls\n", node->NodeName().c_str(), node->OperationName().c_str());
+    }
 
     // determine evaluationNodes from GetEvalCriterionNodes(), ensuring each criterion is only logged once
     std::vector<ComputationNodeBasePtr> evaluationNodes;
@@ -170,9 +175,13 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
 
         if (!evaluationNodes.empty())
         {
-            fprintf(stderr, "\nEvaluation criterion node(s):\n");
+            fprintf(stderr, "\n");
+            LOGPRINTF(stderr, "Evaluation criterion node(s):\n");
+            fprintf(stderr, "\n");
             for (const auto& node : evaluationNodes)
-                fprintf(stderr, "\t%ls = %ls\n", node->NodeName().c_str(), node->OperationName().c_str());
+            {
+                LOGPRINTF(stderr, "\t%ls = %ls\n", node->NodeName().c_str(), node->OperationName().c_str());
+            }
         }
     }
 
@@ -389,8 +398,8 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
 
         if (learnRatePerSample < m_minLearnRate)
         {
-            fprintf(stderr, "Learn Rate Per Sample for Epoch[%d] = %.8g is less than minLearnRate %.8g. Training complete.\n",
-                    i + 1, learnRatePerSample, m_minLearnRate);
+            LOGPRINTF(stderr, "Learn Rate Per Sample for Epoch[%d] = %.8g is less than minLearnRate %.8g. Training complete.\n",
+                      i + 1, learnRatePerSample, m_minLearnRate);
             if (m_autoLearnRateSearchType != LearningRateSearchAlgorithm::None)
             {
                 // In case of parallel training only the main node should we saving the model to prevent
@@ -440,8 +449,9 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
         double momentumAsTimeConstant = momentumPerSample == 0.0 ? 0.0
                                                                  : momentumPerSample >= 1.0 ? 0.0
                                                                                             : -1.0 / log(momentumPerSample);
-        fprintf(stderr, "\nStarting Epoch %d: learning rate per sample = %f  effective momentum = %f  momentum as time constant = %.1f samples\n",
-                i + 1, learnRatePerSample, MomentumPerMB(momentumPerSample, actualMinibatchSize), momentumAsTimeConstant);
+        fprintf(stderr, "\n");
+        LOGPRINTF(stderr, "Starting Epoch %d: learning rate per sample = %f  effective momentum = %f  momentum as time constant = %.1f samples\n",
+                  i + 1, learnRatePerSample, MomentumPerMB(momentumPerSample, actualMinibatchSize), momentumAsTimeConstant);
 
         TrainOneEpoch(net,
                       refNet,
@@ -473,9 +483,9 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
             lrControlCriterion = epochCriterion;
         }
 
-        fprintf(stderr,
-                "Finished Epoch[%2d of %d]: [Training Set] TrainLossPerSample = %.8g; TotalSamplesSeen = %d; ",
-                i + 1, (int)m_maxEpochs, epochCriterion, (int)totalSamplesSeen);
+        LOGPRINTF(stderr,
+                  "Finished Epoch[%2d of %d]: [Training Set] TrainLossPerSample = %.8g; TotalSamplesSeen = %d; ",
+                  i + 1, (int)m_maxEpochs, epochCriterion, (int)totalSamplesSeen);
         m_lastFinishedEpochTrainLoss = epochCriterion;
         if (epochEvalErrors.size() == 0) // no eval criterion, only train criterion itself
         {
@@ -501,13 +511,13 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
                     learnRatePerSample, epochTime);
 
             // TODO: why these extra log messages here and not for 1 eval criterion?
-            fprintf(stderr, "Finished Epoch[%2d of %d]: Criterion Node [%ls] Per Sample = %.8g\n",
-                    i + 1, (int) m_maxEpochs, criterionNodes[0]->NodeName().c_str(), epochCriterion);
+            LOGPRINTF(stderr, "Finished Epoch[%2d of %d]: Criterion Node [%ls] Per Sample = %.8g\n",
+                      i + 1, (int) m_maxEpochs, criterionNodes[0]->NodeName().c_str(), epochCriterion);
 
             for (size_t j = 0; j < epochEvalErrors.size(); j++)
             {
-                fprintf(stderr, "Finished Epoch[%2d of %d]: Evaluation Node [%ls] Per Sample = %.8g\n",
-                        i + 1, (int) m_maxEpochs, evalNodeNames[j].c_str(), epochEvalErrors[j]);
+                LOGPRINTF(stderr, "Finished Epoch[%2d of %d]: Evaluation Node [%ls] Per Sample = %.8g\n",
+                          i + 1, (int) m_maxEpochs, evalNodeNames[j].c_str(), epochEvalErrors[j]);
             }
         }
 
@@ -526,7 +536,7 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
 
                 // BUGBUG: We should not use the training MB size. The training MB size is constrained by both convergence and memory. Eval is only constrained by memory.
             vector<double> vScore = evalforvalidation.Evaluate(validationSetDataReader, cvSetTrainAndEvalNodes, m_mbSize[i]);
-            fprintf(stderr, "Finished Epoch[%2d of %d]: [Validation Set] TrainLossPerSample = %.8g", i + 1, (int) m_maxEpochs, vScore[0]);
+            LOGPRINTF(stderr, "Finished Epoch[%2d of %d]: [Validation Set] TrainLossPerSample = %.8g", i + 1, (int) m_maxEpochs, vScore[0]);
             if (vScore.size() > 1)
             {
                 fprintf(stderr, "; EvalErrPerSample = %.8g", vScore[1]);
@@ -575,7 +585,7 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
                 if (m_loadBestModel)
                 {
                     auto bestModelPath = GetModelNameForEpoch(i - m_learnRateAdjustInterval);
-                    fprintf(stderr, "Loading previous model with best training-criterion value: %ls.\n", bestModelPath.c_str());
+                    LOGPRINTF(stderr, "Loading previous model with best training-criterion value: %ls.\n", bestModelPath.c_str());
                     net->RereadPersistableParameters<ElemType>(bestModelPath);
                     LoadCheckPointInfo(i - m_learnRateAdjustInterval,
                                        /*out*/ totalSamplesSeen,
@@ -604,7 +614,7 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
                         if ((m_mpi == nullptr) || m_mpi->IsMainNode())
                             net->Save(GetModelNameForEpoch(i, true));
 
-                        fprintf(stderr, "Finished training and saved final model\n\n");
+                        LOGPRINTF(stderr, "Finished training and saved final model\n\n");
                         break;
                     }
                 }
@@ -612,7 +622,7 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
                 if (learnRateReduced)
                 {
                     learnRatePerSample *= m_learnRateDecreaseFactor;
-                    fprintf(stderr, "learnRatePerSample reduced to %.8g\n", learnRatePerSample);
+                    LOGPRINTF(stderr, "learnRatePerSample reduced to %.8g\n", learnRatePerSample);
                 }
             }
             else
@@ -623,13 +633,13 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
                 {
 
                     learnRatePerSample *= m_learnRateDecreaseFactor;
-                    fprintf(stderr, "learnRatePerSample reduced to %.8g\n", learnRatePerSample);
+                    LOGPRINTF(stderr, "learnRatePerSample reduced to %.8g\n", learnRatePerSample);
                 }
                 else if (prevCriterion - avgCriterion > m_increaseLearnRateIfImproveMoreThan * prevCriterion &&
                          prevCriterion != std::numeric_limits<double>::infinity())
                 {
                     learnRatePerSample *= m_learnRateIncreaseFactor;
-                    fprintf(stderr, "learnRatePerSample increased to %.8g\n", learnRatePerSample);
+                    LOGPRINTF(stderr, "learnRatePerSample increased to %.8g\n", learnRatePerSample);
                 }
             }
         }
@@ -659,7 +669,7 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
         {
             SaveCheckPointInfo(i, totalSamplesSeen, learnRatePerSample, smoothedGradients, prevCriterion, chosenMinibatchSize);
             auto modelName = GetModelNameForEpoch(i);
-            fprintf(stderr, "SGD: Saving checkpoint model '%ls'\n", modelName.c_str());
+            LOGPRINTF(stderr, "SGD: Saving checkpoint model '%ls'\n", modelName.c_str());
             net->Save(modelName);
             if (!m_keepCheckPointFiles)
             {
@@ -684,8 +694,8 @@ void SGD<ElemType>::TrainOrAdaptModel(int startEpoch, ComputationNetworkPtr net,
 
         if (learnRatePerSample < 1e-12)
         {
-            fprintf(stderr, "learnRate per sample is reduced to %.8g which is below 1e-12. stop training.\n",
-                    learnRatePerSample);
+            LOGPRINTF(stderr, "learnRate per sample is reduced to %.8g which is below 1e-12. stop training.\n",
+                      learnRatePerSample);
         }
     }
     // --- END OF MAIN EPOCH LOOP
@@ -812,6 +822,7 @@ size_t SGD<ElemType>::TrainOneEpoch(ComputationNetworkPtr net,
     // Sub-minibatching is used if a single minibatch is too large to fit into GPU RAM.
     DataReaderHelpers::SubminibatchDispatcher<ElemType> smbDispatcher;
     size_t numSubminibatchesNeeded = DataReaderHelpers::GetNumSubminibatchesNeeded<ElemType>(trainSetDataReader, m_maxSamplesInRAM, m_numSubminiBatches, tunedMBSize);
+
     // this is non-trivial, we need a manager object to handle this
     if (numSubminibatchesNeeded > 1)
         smbDispatcher.Init(net, learnableNodes, criterionNodes, evaluationNodes);
@@ -824,26 +835,30 @@ size_t SGD<ElemType>::TrainOneEpoch(ComputationNetworkPtr net,
     // TODO: move the two-forward-pass support out of the reader, make a first-class citizen.
     AttemptUtteranceDerivativeFeatures(net, trainSetDataReader, featureNodes, inputMatrices);
 
-    fprintf(stderr, "\nStarting minibatch loop");
+    fprintf(stderr, "\n");
+    LOGPRINTF(stderr, "Starting minibatch loop");
     if (useGradientAggregation)
     {
         fprintf(stderr, ", DataParallelSGD training (MyRank = %d, NumNodes = %d, NumGradientBits = %d)",
                 (int) m_mpi->CurrentNodeRank(), (int) m_mpi->NumNodesInUse(), (int) m_numGradientBits);
+
         if (m_bufferedAsyncGradientAggregation)
         {
             fprintf(stderr, ", BufferedAsyncGradientAggregation is ENABLED");
         }
     }
+
     if (useDistributedMBReading)
     {
         fprintf(stderr, ", distributed reading is ENABLED");
     }
+
     if (numSubminibatchesNeeded > 1)
     {
         if (m_maxSamplesInRAM < SIZE_MAX)
-            fprintf(stderr, ", with maximum %d samples in RAM", (int) m_maxSamplesInRAM);
+            fprintf(stderr, ", with maximum %d samples in RAM", (int)m_maxSamplesInRAM);
         else
-            fprintf(stderr, ", with %d subminibatch", (int) numSubminibatchesNeeded);
+            fprintf(stderr, ", with %d subminibatch", (int)numSubminibatchesNeeded);
     }
     fprintf(stderr, ".\n");
 
@@ -1103,7 +1118,7 @@ size_t SGD<ElemType>::TrainOneEpoch(ComputationNetworkPtr net,
                 // progress tracing for regular log
                 string formatString = "%s Epoch[%2d of %d]-Minibatch[%4d-%4d, %2." + std::to_string(mbProgNumPrecision) + "f%%]: SamplesSeen = %d; TrainLossPerSample = " +
                                       GeneratePaddedFloatOrExpFormat(11, 8, trainLossPerSample) + "; ";
-                SGDTrace(stderr, formatString.c_str(),
+                SGDTrace(stderr, true, formatString.c_str(),
                          prefixMsg.c_str(), epochNumber + 1, m_maxEpochs, numMBsRun - m_numMBsToShowResult + 1,
                          numMBsRun, mbProg * 100, numSamplesLastMBs, trainLossPerSample);
             }
@@ -1113,7 +1128,7 @@ size_t SGD<ElemType>::TrainOneEpoch(ComputationNetworkPtr net,
 
                 string formatString = "%s Epoch[%2d of %d]-Minibatch[%4d-%4d]: SamplesSeen = %d; TrainLossPerSample = " +
                                       GeneratePaddedFloatOrExpFormat(11, 8, trainLossPerSample) + "; ";
-                SGDTrace(stderr, formatString.c_str(),
+                SGDTrace(stderr, true, formatString.c_str(),
                          prefixMsg.c_str(), epochNumber + 1, m_maxEpochs, numMBsRun - m_numMBsToShowResult + 1,
                          numMBsRun, numSamplesLastMBs, trainLossPerSample);
                 m_maxComputedEpochSize = numMBsRun * numSamplesLastMBs / m_numMBsToShowResult;
@@ -1124,11 +1139,11 @@ size_t SGD<ElemType>::TrainOneEpoch(ComputationNetworkPtr net,
             {
                 evalError = (epochEvalErrors[i] - epochEvalErrorsLastMBs[i]) / numSamplesLastMBs;
                 string formatString = "EvalErr[%lu]PerSample = " + GeneratePaddedFloatOrExpFormat(0, 8, evalError) + "; ";
-                SGDTrace(stderr, formatString.c_str(), i, evalError);
+                SGDTrace(stderr, false, formatString.c_str(), i, evalError);
             }
 
             string formatString = "TotalTime = " + GeneratePaddedFloatOrExpFormat(0, 4, totalTimeInMBs) + "s; SamplesPerSecond = %.1f\n";
-            SGDTrace(stderr, formatString.c_str(), totalTimeInMBs, numSamplesLastMBs / totalTimeInMBs);
+            SGDTrace(stderr, false, formatString.c_str(), totalTimeInMBs, numSamplesLastMBs / totalTimeInMBs);
 
             // progress tracing for compute cluster management
             if (wasProgressPrinted)
@@ -1287,13 +1302,16 @@ bool SGD<ElemType>::PreCompute(ComputationNetworkPtr net,
 
     if (nodes.size() == 0)
     {
-        fprintf(stderr, "No PreCompute nodes found, skipping PreCompute step.\n");
+        LOGPRINTF(stderr, "No PreCompute nodes found, skipping PreCompute step.\n");
         return false;
     }
 
-    fprintf(stderr, "\nPrecomputing --> %lu PreCompute nodes found.\n\n", nodes.size());
+    fprintf(stderr, "\n");
+    LOGPRINTF(stderr, "Precomputing --> %lu PreCompute nodes found.\n\n", nodes.size());
     for (const auto & node : nodes)
-        fprintf(stderr, "\tNodeName: %ls\n", (node->NodeName()).c_str());
+    {
+        LOGPRINTF(stderr, "\tNodeName: %ls\n", (node->NodeName()).c_str());
+    }
 
     // compute
     ScopedNetworkOperationMode modeGuard(net, NetworkOperationMode::preComputing);
@@ -1328,9 +1346,12 @@ bool SGD<ElemType>::PreCompute(ComputationNetworkPtr net,
 
     // finalize
     for (auto & node : nodes)
+    {
         dynamic_pointer_cast<IPreComputeNode>(node)->MarkComputed(true /*done accumulating*/);
+    }
 
-    fprintf(stderr, "\nPrecomputing --> Completed.\n\n");
+    fprintf(stderr, "\n");
+    LOGPRINTF(stderr, "Precomputing --> Completed.\n\n");
 
     return true;
 }
@@ -1490,8 +1511,8 @@ double SGD<ElemType>::SearchForBestLearnRate(ComputationNetworkPtr net,
         bestLearnRatePerSample = (leftCriterion < rightCriterion) ? leftLearnRatePerSample : rightLearnRatePerSample;
     }
 
-    fprintf(stderr, "Best Learn Rate Per Sample for Epoch[%d] = %.10g  baseCriterion=%.10g\n",
-            epochNumber + 1, bestLearnRatePerSample, baseCriterion);
+    LOGPRINTF(stderr, "Best Learn Rate Per Sample for Epoch[%d] = %.10g  baseCriterion=%.10g\n",
+              epochNumber + 1, bestLearnRatePerSample, baseCriterion);
 
     return bestLearnRatePerSample;
 }
@@ -1542,8 +1563,8 @@ size_t SGD<ElemType>::AdaptiveMinibatchSizing(ComputationNetworkPtr net,
     if (epochNumber < 2 && m_prevChosenMinibatchSize != 0)
     {
         // newly started training: any previous MB size stored in the model is to be ignored
-        fprintf(stderr, "before epoch .2, previous minibatchSize %zd is "
-                        "considered invalid -> resetting\n",
+        LOGPRINTF(stderr, "before epoch .2, previous minibatchSize %zd is "
+                  "considered invalid -> resetting\n",
                 m_prevChosenMinibatchSize);
         m_prevChosenMinibatchSize = 0;
     }
@@ -1553,9 +1574,9 @@ size_t SGD<ElemType>::AdaptiveMinibatchSizing(ComputationNetworkPtr net,
         (epochNumber + 1) > m_minibatchSizeTuningFrequency &&
         (epochNumber + 1) % m_minibatchSizeTuningFrequency != 0)
     {
-        fprintf(stderr, "AdaptiveMinibatchSearch: Search for a better minibatchSize "
-                        "in epoch %d skipped, keeping minibatchSize of %zd\n",
-                epochNumber + 1, m_prevChosenMinibatchSize);
+        LOGPRINTF(stderr, "AdaptiveMinibatchSearch: Search for a better minibatchSize "
+                  "in epoch %d skipped, keeping minibatchSize of %zd\n",
+                  epochNumber + 1, m_prevChosenMinibatchSize);
         chosenMinibatchSize = m_prevChosenMinibatchSize;
     }
     else
@@ -1565,9 +1586,9 @@ size_t SGD<ElemType>::AdaptiveMinibatchSizing(ComputationNetworkPtr net,
             // if m_prevChosenMinibatchSize (the chosen minibatch size for the previous epoch) div 2
             // is higher than initialMinibatchSize (the minibatch size we start with for this epoch),
             // then start the search with m_prevChosenMinibatchSize/2 instead of initialMinibatchSize.
-            fprintf(stderr, "AdaptiveMinibatchSearch: Limiting minMinibatchSize to "
-                            "largest of previous minibatchSize = (%d / 2) or %d\n",
-                    (int) m_prevChosenMinibatchSize, (int) minMinibatchSize);
+            LOGPRINTF(stderr, "AdaptiveMinibatchSearch: Limiting minMinibatchSize to "
+                      "largest of previous minibatchSize = (%d / 2) or %d\n",
+                      (int) m_prevChosenMinibatchSize, (int) minMinibatchSize);
             minMinibatchSize = max(minMinibatchSize, m_prevChosenMinibatchSize / 2);
         }
 
@@ -1578,8 +1599,8 @@ size_t SGD<ElemType>::AdaptiveMinibatchSizing(ComputationNetworkPtr net,
         {
             assert(m_prevChosenMinibatchSize >= chosenMinibatchSize);
 
-            fprintf(stderr, "AdaptiveMinibatchSearch: Limiting maxMinibatchSize to "
-                            "previous minibatchSize %zd*2\n",
+            LOGPRINTF(stderr, "AdaptiveMinibatchSearch: Limiting maxMinibatchSize to "
+                      "previous minibatchSize %zd*2\n",
                     m_prevChosenMinibatchSize);
             maxMinibatchSize = min(maxMinibatchSize, m_prevChosenMinibatchSize * 2);
         }
@@ -1647,8 +1668,9 @@ size_t SGD<ElemType>::SearchForBestMinibatchSize(ComputationNetworkPtr net,
         // round mbsize to something meaningful
         trialMinibatchSize = RoundToMultipleOf64(trialMinibatchSizeFloat);
 
-        fprintf(stderr, "\nAdaptiveMinibatchSearch: Evaluating trial minibatchSize=%zd out of range %zd..%zd ...\n\n",
-                trialMinibatchSize, RoundToMultipleOf64(minMinibatchSize), RoundToMultipleOf64(maxMinibatchSize));
+        fprintf(stderr, "\n");
+        LOGPRINTF(stderr, "AdaptiveMinibatchSearch: Evaluating trial minibatchSize=%zd out of range %zd..%zd ...\n\n",
+                  trialMinibatchSize, RoundToMultipleOf64(minMinibatchSize), RoundToMultipleOf64(maxMinibatchSize));
 
         size_t totalSamplesSeen;
         std::vector<double> epochEvalErrors(evaluationNodes.size(), std::numeric_limits<double>::infinity());
@@ -1675,7 +1697,7 @@ size_t SGD<ElemType>::SearchForBestMinibatchSize(ComputationNetworkPtr net,
             lastTriedTrialEpochCriterion = baseCriterion;
             isFirstIteration = false;
 
-            fprintf(stderr, "AdaptiveMinibatchSearch: Computed BaseCriterion %.10g\n", baseCriterion);
+            LOGPRINTF(stderr, "AdaptiveMinibatchSearch: Computed BaseCriterion %.10g\n", baseCriterion);
         }
         else if (!std::isnan(epochCriterion) &&
                  (epochCriterion > (baseCriterion * (1.0 + (m_minibatchSearchCriterionErrorMargin / 100.0)))))
@@ -1692,15 +1714,15 @@ size_t SGD<ElemType>::SearchForBestMinibatchSize(ComputationNetworkPtr net,
             lastTriedTrialEpochCriterion = epochCriterion;
             if (trialMinibatchSizeFloat * minibatchSizeTuningFactor <= maxMinibatchSize)
             {
-                fprintf(stderr, "AdaptiveMinibatchSearch: Keep searching... "
-                                "EpochCriterion = %.10g vs BaseCriterion = %.10g\n",
-                        epochCriterion, baseCriterion);
+                LOGPRINTF(stderr, "AdaptiveMinibatchSearch: Keep searching... "
+                          "EpochCriterion = %.10g vs BaseCriterion = %.10g\n",
+                          epochCriterion, baseCriterion);
             }
         }
     }
-    fprintf(stderr, "AdaptiveMinibatchSearch: Search successful!!! Chose new minibatchSize of %d. "
-                    "EpochCriterion = %.10g vs BaseCriterion = %.10g\n\n",
-            (int) lastTriedTrialMinibatchSize, lastTriedTrialEpochCriterion, baseCriterion);
+    LOGPRINTF(stderr, "AdaptiveMinibatchSearch: Search successful!!! Chose new minibatchSize of %d. "
+              "EpochCriterion = %.10g vs BaseCriterion = %.10g\n\n",
+              (int) lastTriedTrialMinibatchSize, lastTriedTrialEpochCriterion, baseCriterion);
 
     return lastTriedTrialMinibatchSize;
 }
@@ -1732,18 +1754,18 @@ void SGD<ElemType>::TrainOneMiniEpochAndReloadModel(ComputationNetworkPtr net,
                   /*out*/ epochCriterion, /*out*/ epochEvalErrors, /*out*/ totalSamplesSeen,
                   prefixMsg);
 
-    fprintf(stderr, "Finished Mini-Epoch For LearnRate Selection: TrainLossPerSample = %.8g;", epochCriterion);
+    LOGPRINTF(stderr, "Finished Mini-Epoch For LearnRate Selection: TrainLossPerSample = %.8g;", epochCriterion);
 
     if (epochEvalErrors.size() == 1)
-        fprintf(stderr, "EvalErrPerSample = %.8g; AvgLearningRatePerSample = %.8g\n", epochEvalErrors[0], learnRatePerSample);
+        LOGPRINTF(stderr, "EvalErrPerSample = %.8g; AvgLearningRatePerSample = %.8g\n", epochEvalErrors[0], learnRatePerSample);
     else
     {
-        fprintf(stderr, "EvalErrPerSample ");
+        LOGPRINTF(stderr, "EvalErrPerSample ");
         for (size_t i = 0; i < epochEvalErrors.size(); i++)
         {
-            fprintf(stderr, "[%lu] = %.8g; ", i, epochEvalErrors[i]);
+            LOGPRINTF(stderr, "[%lu] = %.8g; ", i, epochEvalErrors[i]);
         }
-        fprintf(stderr, "AvgLearningRatePerSample = %.8g\n", learnRatePerSample);
+        LOGPRINTF(stderr, "AvgLearningRatePerSample = %.8g\n", learnRatePerSample);
     }
 
     int baseModelEpoch = epochNumber - 1;
@@ -1813,13 +1835,18 @@ static string GeneratePaddedFloatOrExpFormat(int padSize, int precision, double
 }
 
 template <class ElemType>
-int SGD<ElemType>::SGDTrace(FILE* __restrict __stream, const char* __restrict __format, ...)
+int SGD<ElemType>::SGDTrace(FILE* __restrict __stream, bool isPrependTimestamp, const char* __restrict __format, ...)
 {
     int result = 0;
     if (m_traceLevel > 0)
     {
         va_list args;
         va_start(args, __format);
+        if (isPrependTimestamp)
+        {
+            PREPENDTS(__stream);
+        }
+
         result = vfprintf(__stream, __format, args);
         va_end(args);
     }
@@ -1886,10 +1913,10 @@ template <class ElemType>
     // we use simple linear (instead of log linear) scaling here
     const double momentum = MomentumPerMB(momentumPerSample, actualMBSize);
 #if DUMPOUTPUT
-    fprintf(stderr, "learnRatePerSample=%0.8f, momentum=%0.8f, actualMBSize=%ld\n",
-            learnRatePerSample, momentum, actualMBSize);
-    fprintf(stderr, "sgd->GradUpdateType()=%d, sgd->GradientUpdateNoiseStd()=%0.8f\n",
-            sgd->GradUpdateType(), sgd->GradientUpdateNoiseStd());
+    LOGPRINTF(stderr, "learnRatePerSample=%0.8f, momentum=%0.8f, actualMBSize=%ld\n",
+              learnRatePerSample, momentum, actualMBSize);
+    LOGPRINTF(stderr, "sgd->GradUpdateType()=%d, sgd->GradientUpdateNoiseStd()=%0.8f\n",
+              sgd->GradUpdateType(), sgd->GradientUpdateNoiseStd());
     gradientValues.Print("Gradient Input");
     smoothedGradient.Print("Smoothed Gradient Input");
 #endif
@@ -1976,7 +2003,7 @@ void SGD<ElemType>::UpdateWeights(const ComputationNodeBasePtr& node,
                                   const bool useNesterovMomentum) const
 {
 #if DUMPOUTPUT
-    fprintf(stderr, "Update_%ls\n", node->NodeName().c_str());
+    LOGPRINTF(stderr, "Update_%ls\n", node->NodeName().c_str());
 #endif
     if (!node->IsParameterUpdateRequired())
         LogicError("UpdateWeights() called for a learnable ComputationNode which has m_learningRateMultiplier == 0!");
@@ -2072,7 +2099,7 @@ bool SGD<ElemType>::LoadCheckPointInfo(const size_t epochNumber,
     wstring checkPointFileName = GetCheckPointFileNameForEpoch(int(epochNumber));
     if (!fexists(checkPointFileName.c_str()))
     {
-        fprintf(stderr, "Warning: checkpoint file is missing. learning parameters will be initialized from 0\n");
+        LOGPRINTF(stderr, "Warning: checkpoint file is missing. learning parameters will be initialized from 0\n");
         return false;
     }
 
@@ -2167,7 +2194,7 @@ int SGD<ElemType>::DetermineStartEpoch(const bool makeMode)
         }
     }
     if (firstEpoch == m_maxEpochs)
-        fprintf(stderr, "Final model exists: %ls\n", GetModelNameForEpoch(firstEpoch - 1).c_str());
+        LOGPRINTF(stderr, "Final model exists: %ls\n", GetModelNameForEpoch(firstEpoch - 1).c_str());
 
     return firstEpoch;
 }
@@ -2201,7 +2228,8 @@ bool SGD<ElemType>::GradientCheck(ComputationNetworkPtr net,
             irow = max(0, irow);
             icol = max(0, icol);
 
-            fprintf(stderr, "\n###### d%ls######\n", node->NodeName().c_str());
+            fprintf(stderr, "\n");
+            LOGPRINTF(stderr, "###### d%ls######\n", node->NodeName().c_str());
 
             double eOrg = node->Value()(irow, icol);
             node->Value().TransferToDeviceIfNotThere(net->GetDeviceId(), true);
@@ -2259,8 +2287,9 @@ bool SGD<ElemType>::GradientCheck(ComputationNetworkPtr net,
             bool wrong = (std::isnan(diff) || diff > threshold);
             if (wrong)
             {
-                fprintf(stderr, "\nd%ls Numeric gradient = %e, Error BP gradient = %e\n",
-                        node->NodeName().c_str(), eGradNum, eGradErr);
+                fprintf(stderr, "\n");
+                LOGPRINTF(stderr, "d%ls Numeric gradient = %e, Error BP gradient = %e\n",
+                          node->NodeName().c_str(), eGradNum, eGradErr);
                 sprintf(wstrtmp, "\nd%ls Numeric gradient = %e, Error BP gradient = %e\n",
                         node->NodeName().c_str(), eGradNum, eGradErr);
                 errMsgs.push_back(wstrtmp);
diff --git a/Source/SGDLib/SGD.h b/Source/SGDLib/SGD.h
index c0e503786..77f0a3108 100644
--- a/Source/SGDLib/SGD.h
+++ b/Source/SGDLib/SGD.h
@@ -537,7 +537,7 @@ protected:
     shared_ptr<IMASGD<ElemType>> m_pMASGDHelper;
 
 private:
-    int SGDTrace(FILE* __restrict __stream, const char* __restrict __format, ...);
+    int SGDTrace(FILE* __restrict __stream, bool isPrependTimestamp, const char* __restrict __format, ...);
 };
 
 }}}
diff --git a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.cpu.txt b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.cpu.txt
index f686a0048..114e63a93 100644
--- a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.cpu.txt
+++ b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.cpu.txt
@@ -512,7 +512,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14587123; EvalErr[0]PerSample = 0.06400000; TotalTime = 2.60663s; TotalTimePerSample = 10.42652ms; SamplesPerSecond = 95
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15919599; EvalErrPerSample = 0.0765; AvgLearningRatePerSample = 0.00800000038; EpochTime=104.16469
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/06 12:01:30
 MPI Rank 1: command line options: 
@@ -987,7 +987,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14587123; EvalErr[0]PerSample = 0.06400000; TotalTime = 2.63654s; TotalTimePerSample = 10.54614ms; SamplesPerSecond = 94
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15919599; EvalErrPerSample = 0.0765; AvgLearningRatePerSample = 0.00800000038; EpochTime=104.15309
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/06 12:01:30
 MPI Rank 2: command line options: 
@@ -1462,7 +1462,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14587123; EvalErr[0]PerSample = 0.06400000; TotalTime = 2.61322s; TotalTimePerSample = 10.45290ms; SamplesPerSecond = 95
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15919599; EvalErrPerSample = 0.0765; AvgLearningRatePerSample = 0.00800000038; EpochTime=104.16806
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: running on localhost at 2015/10/06 12:01:31
 MPI Rank 3: command line options: 
@@ -1937,5 +1937,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14587123; EvalErr[0]PerSample = 0.06400000; TotalTime = 2.58782s; TotalTimePerSample = 10.35128ms; SamplesPerSecond = 96
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15919599; EvalErrPerSample = 0.0765; AvgLearningRatePerSample = 0.00800000038; EpochTime=104.17161
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.gpu.txt b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.gpu.txt
index e6211f46c..dac2722f0 100644
--- a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.gpu.txt
+++ b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.gpu.txt
@@ -512,7 +512,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.09571s; TotalTimePerSample = 0.38285ms; SamplesPerSecond = 2612
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=3.84576
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/06 11:58:26
 MPI Rank 1: command line options: 
@@ -987,7 +987,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.09571s; TotalTimePerSample = 0.38285ms; SamplesPerSecond = 2612
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=3.845791
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/06 11:58:27
 MPI Rank 2: command line options: 
@@ -1462,7 +1462,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.09571s; TotalTimePerSample = 0.38284ms; SamplesPerSecond = 2612
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=3.845644
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: running on localhost at 2015/10/06 11:58:27
 MPI Rank 3: command line options: 
@@ -1937,5 +1937,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.09571s; TotalTimePerSample = 0.38284ms; SamplesPerSecond = 2612
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=3.845718
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.cpu.txt b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.cpu.txt
index 9d04585f9..b39194776 100644
--- a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.cpu.txt
@@ -518,7 +518,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624377; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.09126s; TotalTimePerSample = 0.36503ms; SamplesPerSecond = 2739
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.159072; EvalErrPerSample = 0.0774; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.243167
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1003,7 +1003,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624377; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.09356s; TotalTimePerSample = 0.37426ms; SamplesPerSecond = 2671
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.159072; EvalErrPerSample = 0.0774; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.24663
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1488,7 +1488,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624377; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.09356s; TotalTimePerSample = 0.37426ms; SamplesPerSecond = 2671
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.159072; EvalErrPerSample = 0.0774; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.246647
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -1973,5 +1973,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624377; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.09126s; TotalTimePerSample = 0.36503ms; SamplesPerSecond = 2739
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.159072; EvalErrPerSample = 0.0774; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.243121
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.gpu.txt b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.gpu.txt
index dd5a59c56..85c4aa1ef 100644
--- a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.gpu.txt
@@ -518,7 +518,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.13598s; TotalTimePerSample = 0.54394ms; SamplesPerSecond = 1838
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.509512
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1003,7 +1003,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.13600s; TotalTimePerSample = 0.54401ms; SamplesPerSecond = 1838
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.509397
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1488,7 +1488,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.13601s; TotalTimePerSample = 0.54403ms; SamplesPerSecond = 1838
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.509323
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -1973,5 +1973,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.13601s; TotalTimePerSample = 0.54403ms; SamplesPerSecond = 1838
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.509212
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.cpu.txt b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.cpu.txt
index 37b880adb..db05e1b81 100644
--- a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.cpu.txt
+++ b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.cpu.txt
@@ -496,7 +496,7 @@ MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20657080; EvalErr[0]PerSample = 0.11600000; TotalTime = 5.56086s; TotalTimePerSample = 22.24344ms; SamplesPerSecond = 44
 MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14566553; EvalErr[0]PerSample = 0.06400000; TotalTime = 5.57143s; TotalTimePerSample = 22.28572ms; SamplesPerSecond = 44
 MPI Rank 0: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15888368; EvalErrPerSample = 0.076499999; AvgLearningRatePerSample = 0.00800000038; EpochTime=222.08354
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/08/25 20:50:23
 MPI Rank 1: command line options: 
@@ -955,7 +955,7 @@ MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20657080; EvalErr[0]PerSample = 0.11600000; TotalTime = 5.56099s; TotalTimePerSample = 22.24397ms; SamplesPerSecond = 44
 MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14566553; EvalErr[0]PerSample = 0.06400000; TotalTime = 5.57339s; TotalTimePerSample = 22.29357ms; SamplesPerSecond = 44
 MPI Rank 1: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15888368; EvalErrPerSample = 0.076499999; AvgLearningRatePerSample = 0.00800000038; EpochTime=222.07455
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/08/25 20:50:24
 MPI Rank 2: command line options: 
@@ -1414,7 +1414,7 @@ MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20657080; EvalErr[0]PerSample = 0.11600000; TotalTime = 5.56393s; TotalTimePerSample = 22.25572ms; SamplesPerSecond = 44
 MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14566553; EvalErr[0]PerSample = 0.06400000; TotalTime = 5.57187s; TotalTimePerSample = 22.28747ms; SamplesPerSecond = 44
 MPI Rank 2: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15888368; EvalErrPerSample = 0.076499999; AvgLearningRatePerSample = 0.00800000038; EpochTime=222.08799
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: running on localhost at 2015/08/25 20:50:24
 MPI Rank 3: command line options: 
@@ -1873,5 +1873,5 @@ MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20657080; EvalErr[0]PerSample = 0.11600000; TotalTime = 5.54955s; TotalTimePerSample = 22.19822ms; SamplesPerSecond = 45
 MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14566553; EvalErr[0]PerSample = 0.06400000; TotalTime = 5.58100s; TotalTimePerSample = 22.32401ms; SamplesPerSecond = 44
 MPI Rank 3: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15888368; EvalErrPerSample = 0.076499999; AvgLearningRatePerSample = 0.00800000038; EpochTime=222.07455
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.gpu.txt b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.gpu.txt
index 61f4a7554..4615d227f 100644
--- a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.gpu.txt
+++ b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.gpu.txt
@@ -631,7 +631,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12570s; TotalTimePerSample = 0.50281ms; SamplesPerSecond = 1988
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.018144
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/24 12:44:54
 MPI Rank 1: command line: 
@@ -1225,7 +1225,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12566s; TotalTimePerSample = 0.50264ms; SamplesPerSecond = 1989
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.01855
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/24 12:44:54
 MPI Rank 2: command line: 
@@ -1819,7 +1819,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12566s; TotalTimePerSample = 0.50262ms; SamplesPerSecond = 1989
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.018583
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: running on localhost at 2015/10/24 12:44:55
 MPI Rank 3: command line: 
@@ -2413,5 +2413,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12570s; TotalTimePerSample = 0.50282ms; SamplesPerSecond = 1988
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.018182
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.cpu.txt b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.cpu.txt
index 135d6ff42..b047b3277 100644
--- a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.cpu.txt
@@ -502,7 +502,7 @@ MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20619434; EvalErr[0]PerSample = 0.11200000; TotalTime = 0.12883s; TotalTimePerSample = 0.51532ms; SamplesPerSecond = 1940
 MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624365; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.12283s; TotalTimePerSample = 0.49132ms; SamplesPerSecond = 2035
 MPI Rank 0: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15907188; EvalErrPerSample = 0.077399999; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.887408
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20150825180808.217636\ParallelTraining\NoQuantization_SinglePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -971,7 +971,7 @@ MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20619434; EvalErr[0]PerSample = 0.11200000; TotalTime = 0.12881s; TotalTimePerSample = 0.51526ms; SamplesPerSecond = 1940
 MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624365; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.12287s; TotalTimePerSample = 0.49148ms; SamplesPerSecond = 2034
 MPI Rank 1: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15907188; EvalErrPerSample = 0.077399999; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.887381
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20150825180808.217636\ParallelTraining\NoQuantization_SinglePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1440,7 +1440,7 @@ MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20619434; EvalErr[0]PerSample = 0.11200000; TotalTime = 0.12898s; TotalTimePerSample = 0.51592ms; SamplesPerSecond = 1938
 MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624365; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.12295s; TotalTimePerSample = 0.49182ms; SamplesPerSecond = 2033
 MPI Rank 2: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15907188; EvalErrPerSample = 0.077399999; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.887366
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20150825180808.217636\ParallelTraining\NoQuantization_SinglePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -1909,5 +1909,5 @@ MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20619434; EvalErr[0]PerSample = 0.11200000; TotalTime = 0.12882s; TotalTimePerSample = 0.51529ms; SamplesPerSecond = 1940
 MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624365; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.12279s; TotalTimePerSample = 0.49116ms; SamplesPerSecond = 2036
 MPI Rank 3: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15907188; EvalErrPerSample = 0.077399999; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.887371
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.gpu.txt b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.gpu.txt
index e7c567b18..8ec779c8f 100644
--- a/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/CNTKTextFormatReader/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.gpu.txt
@@ -636,7 +636,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12520s; TotalTimePerSample = 0.50081ms; SamplesPerSecond = 1996
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.931563
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151024134937.188247\ParallelTraining\NoQuantization_SinglePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1239,7 +1239,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12625s; TotalTimePerSample = 0.50498ms; SamplesPerSecond = 1980
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.931591
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151024134937.188247\ParallelTraining\NoQuantization_SinglePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1842,7 +1842,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12606s; TotalTimePerSample = 0.50426ms; SamplesPerSecond = 1983
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.931381
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151024134937.188247\ParallelTraining\NoQuantization_SinglePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -2445,5 +2445,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12588s; TotalTimePerSample = 0.50353ms; SamplesPerSecond = 1985
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.931393
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.debug.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.debug.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.release.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.release.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.linux.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.debug.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.debug.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.release.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.release.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/baseline.windows.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/testcases.yml b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/testcases.yml
index e52344bfb..713182b70 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/testcases.yml
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/01_OneHidden/testcases.yml
@@ -9,5 +9,5 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.debug.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.debug.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.release.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.release.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.linux.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.debug.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.debug.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.release.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.release.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/baseline.windows.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/testcases.yml b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/testcases.yml
index f7d5751e8..485f09752 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/testcases.yml
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution/testcases.yml
@@ -11,5 +11,5 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.debug.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.debug.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.release.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.release.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.linux.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.debug.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.debug.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.release.cpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.release.gpu.txt b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/baseline.windows.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/testcases.yml b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/testcases.yml
index 68b7fed82..3ad32557c 100644
--- a/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/testcases.yml
+++ b/Tests/EndToEndTests/Examples/Image/MNIST/03_ConvBatchNorm/testcases.yml
@@ -9,5 +9,5 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.debug.cpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.debug.gpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.release.cpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.release.gpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.linux.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.debug.cpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.debug.gpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.release.cpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.release.gpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/baseline.windows.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/testcases.yml b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/testcases.yml
index 3e1053097..bef057ef1 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/testcases.yml
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/MultiGpu/testcases.yml
@@ -9,5 +9,5 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.debug.cpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.debug.gpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.release.cpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.release.gpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.linux.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.debug.cpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.debug.gpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.release.cpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.release.gpu.txt b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/baseline.windows.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/testcases.yml b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/testcases.yml
index 200bdabb7..2a89172fa 100644
--- a/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/testcases.yml
+++ b/Tests/EndToEndTests/Examples/Other/Simple2d/Simple/testcases.yml
@@ -9,5 +9,5 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.debug.cpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.debug.gpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.release.cpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.release.gpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.linux.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.debug.cpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.debug.gpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.release.cpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.release.gpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/baseline.windows.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/testcases.yml b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/testcases.yml
index 03ab69821..3a8ae7f09 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/testcases.yml
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/FeedForward/testcases.yml
@@ -9,5 +9,5 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.debug.cpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.debug.gpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.release.cpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.release.gpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.linux.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.debug.cpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.debug.gpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.release.cpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.release.gpu.txt b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/baseline.windows.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/testcases.yml b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/testcases.yml
index 03ab69821..3a8ae7f09 100644
--- a/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/testcases.yml
+++ b/Tests/EndToEndTests/Examples/Speech/AN4/LSTM/testcases.yml
@@ -9,5 +9,5 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
diff --git a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.debug.cpu.txt b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.debug.gpu.txt b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.release.cpu.txt b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.release.gpu.txt b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.linux.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.debug.cpu.txt b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.debug.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.debug.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.debug.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.debug.gpu.txt b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.debug.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.debug.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.debug.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.release.cpu.txt b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.release.cpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.release.cpu.txt
+++ b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.release.cpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.release.gpu.txt b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.release.gpu.txt
index 814710451..284dfdff9 100644
--- a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.release.gpu.txt
+++ b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/baseline.windows.release.gpu.txt
@@ -1 +1 @@
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/testcases.yml b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/testcases.yml
index eadfab345..2d4d1a988 100644
--- a/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/testcases.yml
+++ b/Tests/EndToEndTests/Examples/Text/PennTreebank/RNN/testcases.yml
@@ -9,5 +9,5 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
diff --git a/Tests/EndToEndTests/Image/AlexNet/baseline.linux.debug.gpu.txt b/Tests/EndToEndTests/Image/AlexNet/baseline.linux.debug.gpu.txt
index ebf740566..f670f0100 100644
--- a/Tests/EndToEndTests/Image/AlexNet/baseline.linux.debug.gpu.txt
+++ b/Tests/EndToEndTests/Image/AlexNet/baseline.linux.debug.gpu.txt
@@ -2533,4 +2533,4 @@ evalNodeNames are not specified, using all the default evalnodes and training cr
 Allocating matrices for forward and/or backward propagation.
 Minibatch[1-32]: Samples Seen = 500    Err: ErrorPrediction/Sample = 0.998    errTop5: ErrorPrediction/Sample = 0.992    CE: CrossEntropyWithSoftmax/Sample = 6.9591762    
 Final Results: Minibatch[1-32]: Samples Seen = 500    Err: ErrorPrediction/Sample = 0.998    errTop5: ErrorPrediction/Sample = 0.992    CE: CrossEntropyWithSoftmax/Sample = 6.9591762    Perplexity = 1052.766    
-COMPLETED
\ No newline at end of file
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Image/AlexNet/baseline.linux.release.gpu.txt b/Tests/EndToEndTests/Image/AlexNet/baseline.linux.release.gpu.txt
index 410cebeaf..180170511 100644
--- a/Tests/EndToEndTests/Image/AlexNet/baseline.linux.release.gpu.txt
+++ b/Tests/EndToEndTests/Image/AlexNet/baseline.linux.release.gpu.txt
@@ -1075,9 +1075,11 @@ evalNodeNames are not specified, using all the default evalnodes and training cr
 
 
 Allocating matrices for forward and/or backward propagation.
+
 Minibatch[1-32]: SamplesSeen = 500    Err: ErrorPrediction/Sample = 0.998    errTop5: ErrorPrediction/Sample = 0.996    CE: CrossEntropyWithSoftmax/Sample = 6.9611959    
 Final Results: Minibatch[1-32]: SamplesSeen = 500    Err: ErrorPrediction/Sample = 0.998    errTop5: ErrorPrediction/Sample = 0.996    CE: CrossEntropyWithSoftmax/Sample = 6.9611959    Perplexity = 1054.8943    
 
 Action "test" complete.
 
-COMPLETED
\ No newline at end of file
+__COMPLETED__
+
diff --git a/Tests/EndToEndTests/Image/AlexNet/baseline.windows.debug.gpu.txt b/Tests/EndToEndTests/Image/AlexNet/baseline.windows.debug.gpu.txt
index 837103565..ce3b68629 100644
--- a/Tests/EndToEndTests/Image/AlexNet/baseline.windows.debug.gpu.txt
+++ b/Tests/EndToEndTests/Image/AlexNet/baseline.windows.debug.gpu.txt
@@ -2981,4 +2981,4 @@ CUDA error 11 [c:\tools\cub-1.4.1\cub\device\dispatch/dispatch_radix_sort.cuh, 7
 CUDA error 11 [c:\tools\cub-1.4.1\cub\device\dispatch/dispatch_radix_sort.cuh, 796]: invalid argument
 Minibatch[1-32]: Samples Seen = 500    Err: ErrorPrediction/Sample = 1    errTop5: ErrorPrediction/Sample = 0.992    CE: CrossEntropyWithSoftmax/Sample = 6.9566009    
 Final Results: Minibatch[1-32]: Samples Seen = 500    Err: ErrorPrediction/Sample = 1    errTop5: ErrorPrediction/Sample = 0.992    CE: CrossEntropyWithSoftmax/Sample = 6.9566009    Perplexity = 1050.0582    
-COMPLETED
\ No newline at end of file
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Image/AlexNet/baseline.windows.release.gpu.txt b/Tests/EndToEndTests/Image/AlexNet/baseline.windows.release.gpu.txt
index 7c41a970a..c8ed55786 100644
--- a/Tests/EndToEndTests/Image/AlexNet/baseline.windows.release.gpu.txt
+++ b/Tests/EndToEndTests/Image/AlexNet/baseline.windows.release.gpu.txt
@@ -2533,4 +2533,4 @@ evalNodeNames are not specified, using all the default evalnodes and training cr
 Allocating matrices for forward and/or backward propagation.
 Minibatch[1-32]: Samples Seen = 500    Err: ErrorPrediction/Sample = 1    errTop5: ErrorPrediction/Sample = 0.996    CE: CrossEntropyWithSoftmax/Sample = 6.9640379    
 Final Results: Minibatch[1-32]: Samples Seen = 500    Err: ErrorPrediction/Sample = 1    errTop5: ErrorPrediction/Sample = 0.996    CE: CrossEntropyWithSoftmax/Sample = 6.9640379    Perplexity = 1057.8967    
-COMPLETED
\ No newline at end of file
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Image/AlexNet/testcases.yml b/Tests/EndToEndTests/Image/AlexNet/testcases.yml
index 8cc7a4340..cb80a662c 100644
--- a/Tests/EndToEndTests/Image/AlexNet/testcases.yml
+++ b/Tests/EndToEndTests/Image/AlexNet/testcases.yml
@@ -8,24 +8,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.2%}}
       - EvalErrPerSample = {{float,tolerance=.2%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.2%}}
 
diff --git a/Tests/EndToEndTests/Image/QuickE2E/baseline.linux.debug.gpu.txt b/Tests/EndToEndTests/Image/QuickE2E/baseline.linux.debug.gpu.txt
index 3c5b6f82d..88bc52f36 100644
--- a/Tests/EndToEndTests/Image/QuickE2E/baseline.linux.debug.gpu.txt
+++ b/Tests/EndToEndTests/Image/QuickE2E/baseline.linux.debug.gpu.txt
@@ -913,7 +913,7 @@ already there from last epoch
 RandomOrdering: 21 retries for 100 elements (21.0%) to ensure window condition
 RandomOrdering: recached sequence for seed 0: 38, 46, ...
 Final Results: Minibatch[1-1]: Samples Seen = 100    err: ErrorPrediction/Sample = 0    ce: CrossEntropyWithSoftmax/Sample = 0.30271576    Perplexity = 1.3535297    
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 
@@ -1734,4 +1734,4 @@ already there from last epoch
 RandomOrdering: 21 retries for 100 elements (21.0%) to ensure window condition
 RandomOrdering: recached sequence for seed 0: 38, 46, ...
 Final Results: Minibatch[1-1]: Samples Seen = 100    err: ErrorPrediction/Sample = 0    ce: CrossEntropyWithSoftmax/Sample = 0.31798401    Perplexity = 1.3743543    
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Image/QuickE2E/baseline.linux.release.gpu.txt b/Tests/EndToEndTests/Image/QuickE2E/baseline.linux.release.gpu.txt
index 2dcc47bf6..b2356fea8 100644
--- a/Tests/EndToEndTests/Image/QuickE2E/baseline.linux.release.gpu.txt
+++ b/Tests/EndToEndTests/Image/QuickE2E/baseline.linux.release.gpu.txt
@@ -915,7 +915,7 @@ RandomOrdering: 21 retries for 100 elements (21.0%) to ensure window condition
 RandomOrdering: recached sequence for seed 0: 38, 46, ...
 MBLayout::Init: Resizing m_distanceToStart from 1 x 0 to 100 x 1
 Final Results: Minibatch[1-1]: Samples Seen = 100    err: ErrorPrediction/Sample = 0    ce: CrossEntropyWithSoftmax/Sample = 0.30270519    Perplexity = 1.3535154    
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 -------------------------------------------------------------------
@@ -1737,4 +1737,4 @@ RandomOrdering: 21 retries for 100 elements (21.0%) to ensure window condition
 RandomOrdering: recached sequence for seed 0: 38, 46, ...
 MBLayout::Init: Resizing m_distanceToStart from 1 x 0 to 100 x 1
 Final Results: Minibatch[1-1]: Samples Seen = 100    err: ErrorPrediction/Sample = 0    ce: CrossEntropyWithSoftmax/Sample = 0.31781933    Perplexity = 1.374128    
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.debug.cpu.txt b/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.debug.cpu.txt
index 396954569..e088befb2 100644
--- a/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.debug.cpu.txt
+++ b/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.debug.cpu.txt
@@ -1468,7 +1468,7 @@ already there from last epoch
 randomordering: 11 retries for 100 elements (11.0%) to ensure window condition
 randomordering: recached sequence for seed 0: 15, 33, ...
 Final Results: Minibatch[1-1]: Samples Seen = 100    Err: ErrorPrediction/Sample = 0    CE: CrossEntropyWithSoftmax/Sample = 0.31800278    Perplexity = 1.3743801    
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 -------------------------------------------------------------------
@@ -2844,4 +2844,4 @@ already there from last epoch
 randomordering: 11 retries for 100 elements (11.0%) to ensure window condition
 randomordering: recached sequence for seed 0: 15, 33, ...
 Final Results: Minibatch[1-1]: Samples Seen = 100    Err: ErrorPrediction/Sample = 0    CE: CrossEntropyWithSoftmax/Sample = 0.33039909    Perplexity = 1.3915234    
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.debug.gpu.txt b/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.debug.gpu.txt
index a0f43cc3a..27e5f3435 100644
--- a/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.debug.gpu.txt
+++ b/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.debug.gpu.txt
@@ -910,7 +910,7 @@ already there from last epoch
 RandomOrdering: 11 retries for 100 elements (11.0%) to ensure window condition
 RandomOrdering: recached sequence for seed 0: 15, 33, ...
 Final Results: Minibatch[1-1]: Samples Seen = 100    err: ErrorPrediction/Sample = 0    ce: CrossEntropyWithSoftmax/Sample = 0.29111851    Perplexity = 1.3379231    
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 -------------------------------------------------------------------
@@ -1727,4 +1727,4 @@ already there from last epoch
 RandomOrdering: 11 retries for 100 elements (11.0%) to ensure window condition
 RandomOrdering: recached sequence for seed 0: 15, 33, ...
 Final Results: Minibatch[1-1]: Samples Seen = 100    err: ErrorPrediction/Sample = 0    ce: CrossEntropyWithSoftmax/Sample = 0.30440025    Perplexity = 1.3558116    
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.release.cpu.txt b/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.release.cpu.txt
index aacee7bf8..65a8204e5 100644
--- a/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.release.cpu.txt
+++ b/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.release.cpu.txt
@@ -1410,7 +1410,7 @@ already there from last epoch
 randomordering: 11 retries for 100 elements (11.0%) to ensure window condition
 randomordering: recached sequence for seed 0: 15, 33, ...
 Final Results: Minibatch[1-1]: Samples Seen = 100    Err: ErrorPrediction/Sample = 0    CE: CrossEntropyWithSoftmax/Sample = 0.31800278    Perplexity = 1.3743801    
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 -------------------------------------------------------------------
@@ -2728,4 +2728,4 @@ already there from last epoch
 randomordering: 11 retries for 100 elements (11.0%) to ensure window condition
 randomordering: recached sequence for seed 0: 15, 33, ...
 Final Results: Minibatch[1-1]: Samples Seen = 100    Err: ErrorPrediction/Sample = 0    CE: CrossEntropyWithSoftmax/Sample = 0.33039909    Perplexity = 1.3915234    
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.release.gpu.txt b/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.release.gpu.txt
index c34022c07..ee2e4180a 100644
--- a/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.release.gpu.txt
+++ b/Tests/EndToEndTests/Image/QuickE2E/baseline.windows.release.gpu.txt
@@ -910,7 +910,7 @@ already there from last epoch
 RandomOrdering: 11 retries for 100 elements (11.0%) to ensure window condition
 RandomOrdering: recached sequence for seed 0: 15, 33, ...
 Final Results: Minibatch[1-1]: Samples Seen = 100    err: ErrorPrediction/Sample = 0    ce: CrossEntropyWithSoftmax/Sample = 0.29111847    Perplexity = 1.3379231    
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 -------------------------------------------------------------------
@@ -1727,4 +1727,4 @@ already there from last epoch
 RandomOrdering: 11 retries for 100 elements (11.0%) to ensure window condition
 RandomOrdering: recached sequence for seed 0: 15, 33, ...
 Final Results: Minibatch[1-1]: Samples Seen = 100    err: ErrorPrediction/Sample = 0    ce: CrossEntropyWithSoftmax/Sample = 0.30440022    Perplexity = 1.3558116    
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Image/QuickE2E/testcases.yml b/Tests/EndToEndTests/Image/QuickE2E/testcases.yml
index 5ae4a1ea7..4d57433b0 100644
--- a/Tests/EndToEndTests/Image/QuickE2E/testcases.yml
+++ b/Tests/EndToEndTests/Image/QuickE2E/testcases.yml
@@ -8,24 +8,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.1%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErr[0]PerSample = {{float,tolerance=.1%}}
diff --git a/Tests/EndToEndTests/ModelExport/Model0/baseline.txt b/Tests/EndToEndTests/ModelExport/Model0/baseline.txt
index 909445983..f12ae3a43 100644
--- a/Tests/EndToEndTests/ModelExport/Model0/baseline.txt
+++ b/Tests/EndToEndTests/ModelExport/Model0/baseline.txt
@@ -1,2 +1,2 @@
 Post-processing network complete.
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/ModelExport/Model0/testcases.yml b/Tests/EndToEndTests/ModelExport/Model0/testcases.yml
index cbfcef620..549521a66 100644
--- a/Tests/EndToEndTests/ModelExport/Model0/testcases.yml
+++ b/Tests/EndToEndTests/ModelExport/Model0/testcases.yml
@@ -8,5 +8,5 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
diff --git a/Tests/EndToEndTests/ModelExport/Model1/baseline.txt b/Tests/EndToEndTests/ModelExport/Model1/baseline.txt
index 909445983..f12ae3a43 100644
--- a/Tests/EndToEndTests/ModelExport/Model1/baseline.txt
+++ b/Tests/EndToEndTests/ModelExport/Model1/baseline.txt
@@ -1,2 +1,2 @@
 Post-processing network complete.
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/ModelExport/Model1/testcases.yml b/Tests/EndToEndTests/ModelExport/Model1/testcases.yml
index cbfcef620..549521a66 100644
--- a/Tests/EndToEndTests/ModelExport/Model1/testcases.yml
+++ b/Tests/EndToEndTests/ModelExport/Model1/testcases.yml
@@ -8,5 +8,5 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
diff --git a/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.cpu.txt b/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.cpu.txt
index f686a0048..114e63a93 100644
--- a/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.cpu.txt
+++ b/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.cpu.txt
@@ -512,7 +512,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14587123; EvalErr[0]PerSample = 0.06400000; TotalTime = 2.60663s; TotalTimePerSample = 10.42652ms; SamplesPerSecond = 95
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15919599; EvalErrPerSample = 0.0765; AvgLearningRatePerSample = 0.00800000038; EpochTime=104.16469
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/06 12:01:30
 MPI Rank 1: command line options: 
@@ -987,7 +987,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14587123; EvalErr[0]PerSample = 0.06400000; TotalTime = 2.63654s; TotalTimePerSample = 10.54614ms; SamplesPerSecond = 94
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15919599; EvalErrPerSample = 0.0765; AvgLearningRatePerSample = 0.00800000038; EpochTime=104.15309
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/06 12:01:30
 MPI Rank 2: command line options: 
@@ -1462,7 +1462,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14587123; EvalErr[0]PerSample = 0.06400000; TotalTime = 2.61322s; TotalTimePerSample = 10.45290ms; SamplesPerSecond = 95
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15919599; EvalErrPerSample = 0.0765; AvgLearningRatePerSample = 0.00800000038; EpochTime=104.16806
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: running on localhost at 2015/10/06 12:01:31
 MPI Rank 3: command line options: 
@@ -1937,5 +1937,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14587123; EvalErr[0]PerSample = 0.06400000; TotalTime = 2.58782s; TotalTimePerSample = 10.35128ms; SamplesPerSecond = 96
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15919599; EvalErrPerSample = 0.0765; AvgLearningRatePerSample = 0.00800000038; EpochTime=104.17161
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.gpu.txt b/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.gpu.txt
index e6211f46c..dac2722f0 100644
--- a/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.gpu.txt
+++ b/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.gpu.txt
@@ -512,7 +512,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.09571s; TotalTimePerSample = 0.38285ms; SamplesPerSecond = 2612
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=3.84576
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/06 11:58:26
 MPI Rank 1: command line options: 
@@ -987,7 +987,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.09571s; TotalTimePerSample = 0.38285ms; SamplesPerSecond = 2612
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=3.845791
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/06 11:58:27
 MPI Rank 2: command line options: 
@@ -1462,7 +1462,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.09571s; TotalTimePerSample = 0.38284ms; SamplesPerSecond = 2612
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=3.845644
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: running on localhost at 2015/10/06 11:58:27
 MPI Rank 3: command line options: 
@@ -1937,5 +1937,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.09571s; TotalTimePerSample = 0.38284ms; SamplesPerSecond = 2612
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=3.845718
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.cpu.txt b/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.cpu.txt
index 9d04585f9..b39194776 100644
--- a/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.cpu.txt
@@ -518,7 +518,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624377; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.09126s; TotalTimePerSample = 0.36503ms; SamplesPerSecond = 2739
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.159072; EvalErrPerSample = 0.0774; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.243167
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1003,7 +1003,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624377; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.09356s; TotalTimePerSample = 0.37426ms; SamplesPerSecond = 2671
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.159072; EvalErrPerSample = 0.0774; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.24663
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1488,7 +1488,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624377; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.09356s; TotalTimePerSample = 0.37426ms; SamplesPerSecond = 2671
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.159072; EvalErrPerSample = 0.0774; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.246647
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -1973,5 +1973,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624377; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.09126s; TotalTimePerSample = 0.36503ms; SamplesPerSecond = 2739
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.159072; EvalErrPerSample = 0.0774; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.243121
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.gpu.txt b/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.gpu.txt
index dd5a59c56..85c4aa1ef 100644
--- a/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/ParallelTraining/NoQuantization/DoublePrecision/baseline.windows.gpu.txt
@@ -518,7 +518,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.13598s; TotalTimePerSample = 0.54394ms; SamplesPerSecond = 1838
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.509512
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1003,7 +1003,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.13600s; TotalTimePerSample = 0.54401ms; SamplesPerSecond = 1838
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.509397
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1488,7 +1488,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.13601s; TotalTimePerSample = 0.54403ms; SamplesPerSecond = 1838
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.509323
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006100025.151784\ParallelTraining\NoQuantization_DoublePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -1973,5 +1973,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14604076; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.13601s; TotalTimePerSample = 0.54403ms; SamplesPerSecond = 1838
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15920517; EvalErrPerSample = 0.0766; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.509212
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.cpu.txt b/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.cpu.txt
index 37b880adb..db05e1b81 100644
--- a/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.cpu.txt
+++ b/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.cpu.txt
@@ -496,7 +496,7 @@ MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20657080; EvalErr[0]PerSample = 0.11600000; TotalTime = 5.56086s; TotalTimePerSample = 22.24344ms; SamplesPerSecond = 44
 MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14566553; EvalErr[0]PerSample = 0.06400000; TotalTime = 5.57143s; TotalTimePerSample = 22.28572ms; SamplesPerSecond = 44
 MPI Rank 0: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15888368; EvalErrPerSample = 0.076499999; AvgLearningRatePerSample = 0.00800000038; EpochTime=222.08354
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/08/25 20:50:23
 MPI Rank 1: command line options: 
@@ -955,7 +955,7 @@ MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20657080; EvalErr[0]PerSample = 0.11600000; TotalTime = 5.56099s; TotalTimePerSample = 22.24397ms; SamplesPerSecond = 44
 MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14566553; EvalErr[0]PerSample = 0.06400000; TotalTime = 5.57339s; TotalTimePerSample = 22.29357ms; SamplesPerSecond = 44
 MPI Rank 1: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15888368; EvalErrPerSample = 0.076499999; AvgLearningRatePerSample = 0.00800000038; EpochTime=222.07455
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/08/25 20:50:24
 MPI Rank 2: command line options: 
@@ -1414,7 +1414,7 @@ MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20657080; EvalErr[0]PerSample = 0.11600000; TotalTime = 5.56393s; TotalTimePerSample = 22.25572ms; SamplesPerSecond = 44
 MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14566553; EvalErr[0]PerSample = 0.06400000; TotalTime = 5.57187s; TotalTimePerSample = 22.28747ms; SamplesPerSecond = 44
 MPI Rank 2: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15888368; EvalErrPerSample = 0.076499999; AvgLearningRatePerSample = 0.00800000038; EpochTime=222.08799
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: running on localhost at 2015/08/25 20:50:24
 MPI Rank 3: command line options: 
@@ -1873,5 +1873,5 @@ MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20657080; EvalErr[0]PerSample = 0.11600000; TotalTime = 5.54955s; TotalTimePerSample = 22.19822ms; SamplesPerSecond = 45
 MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14566553; EvalErr[0]PerSample = 0.06400000; TotalTime = 5.58100s; TotalTimePerSample = 22.32401ms; SamplesPerSecond = 44
 MPI Rank 3: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15888368; EvalErrPerSample = 0.076499999; AvgLearningRatePerSample = 0.00800000038; EpochTime=222.07455
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.gpu.txt b/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.gpu.txt
index 61f4a7554..4615d227f 100644
--- a/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.gpu.txt
+++ b/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.gpu.txt
@@ -631,7 +631,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12570s; TotalTimePerSample = 0.50281ms; SamplesPerSecond = 1988
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.018144
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/24 12:44:54
 MPI Rank 1: command line: 
@@ -1225,7 +1225,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12566s; TotalTimePerSample = 0.50264ms; SamplesPerSecond = 1989
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.01855
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/24 12:44:54
 MPI Rank 2: command line: 
@@ -1819,7 +1819,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12566s; TotalTimePerSample = 0.50262ms; SamplesPerSecond = 1989
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.018583
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: running on localhost at 2015/10/24 12:44:55
 MPI Rank 3: command line: 
@@ -2413,5 +2413,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12570s; TotalTimePerSample = 0.50282ms; SamplesPerSecond = 1988
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=5.018182
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.cpu.txt b/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.cpu.txt
index 135d6ff42..b047b3277 100644
--- a/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.cpu.txt
@@ -502,7 +502,7 @@ MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20619434; EvalErr[0]PerSample = 0.11200000; TotalTime = 0.12883s; TotalTimePerSample = 0.51532ms; SamplesPerSecond = 1940
 MPI Rank 0:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624365; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.12283s; TotalTimePerSample = 0.49132ms; SamplesPerSecond = 2035
 MPI Rank 0: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15907188; EvalErrPerSample = 0.077399999; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.887408
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20150825180808.217636\ParallelTraining\NoQuantization_SinglePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -971,7 +971,7 @@ MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20619434; EvalErr[0]PerSample = 0.11200000; TotalTime = 0.12881s; TotalTimePerSample = 0.51526ms; SamplesPerSecond = 1940
 MPI Rank 1:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624365; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.12287s; TotalTimePerSample = 0.49148ms; SamplesPerSecond = 2034
 MPI Rank 1: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15907188; EvalErrPerSample = 0.077399999; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.887381
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20150825180808.217636\ParallelTraining\NoQuantization_SinglePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1440,7 +1440,7 @@ MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20619434; EvalErr[0]PerSample = 0.11200000; TotalTime = 0.12898s; TotalTimePerSample = 0.51592ms; SamplesPerSecond = 1938
 MPI Rank 2:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624365; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.12295s; TotalTimePerSample = 0.49182ms; SamplesPerSecond = 2033
 MPI Rank 2: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15907188; EvalErrPerSample = 0.077399999; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.887366
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20150825180808.217636\ParallelTraining\NoQuantization_SinglePrecision@debug_cpu/stderr_SimpleMultiGPU.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -1909,5 +1909,5 @@ MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 371- 380]: SamplesSeen = 250; TrainLossP
 MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 381- 390]: SamplesSeen = 250; TrainLossPerSample =  0.20619434; EvalErr[0]PerSample = 0.11200000; TotalTime = 0.12882s; TotalTimePerSample = 0.51529ms; SamplesPerSecond = 1940
 MPI Rank 3:  Epoch[ 4 of 10]-Minibatch[ 391- 400]: SamplesSeen = 250; TrainLossPerSample =  0.14624365; EvalErr[0]PerSample = 0.06800000; TotalTime = 0.12279s; TotalTimePerSample = 0.49116ms; SamplesPerSecond = 2036
 MPI Rank 3: Finished Epoch[ 4 of 10]: [Training Set] TrainLossPerSample = 0.15907188; EvalErrPerSample = 0.077399999; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.887371
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.gpu.txt b/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.gpu.txt
index e7c567b18..8ec779c8f 100644
--- a/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/ParallelTraining/NoQuantization/SinglePrecision/baseline.windows.gpu.txt
@@ -636,7 +636,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12520s; TotalTimePerSample = 0.50081ms; SamplesPerSecond = 1996
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.931563
 MPI Rank 0: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151024134937.188247\ParallelTraining\NoQuantization_SinglePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1239,7 +1239,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12625s; TotalTimePerSample = 0.50498ms; SamplesPerSecond = 1980
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.931591
 MPI Rank 1: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151024134937.188247\ParallelTraining\NoQuantization_SinglePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1842,7 +1842,7 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12606s; TotalTimePerSample = 0.50426ms; SamplesPerSecond = 1983
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.931381
 MPI Rank 2: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151024134937.188247\ParallelTraining\NoQuantization_SinglePrecision@debug_gpu/stderr_SimpleMultiGPU.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -2445,5 +2445,5 @@ MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 381- 390 of 400]: SamplesSeen = 250; Trai
 MPI Rank 3:  Epoch[ 4 of 4]-Minibatch[ 391- 400 of 400]: SamplesSeen = 250; TrainLossPerSample =  0.14585245; EvalErr[0]PerSample = 0.06400000; TotalTime = 0.12588s; TotalTimePerSample = 0.50353ms; SamplesPerSecond = 1985
 MPI Rank 3: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.15914931; EvalErrPerSample = 0.0767; AvgLearningRatePerSample = 0.00800000038; EpochTime=4.931393
 MPI Rank 3: CNTKCommandTrainEnd: SimpleMultiGPU
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.cpu.txt b/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.cpu.txt
index 82fdb6138..dfea1ce77 100644
--- a/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.cpu.txt
@@ -3265,4 +3265,4 @@ Starting minibatch loop.
  Epoch[ 4 of 4]-Minibatch[ 151- 160 of 160]: SamplesSeen = 5120; TrainLossPerSample =  0.97683563; EvalErr[0]PerSample = 0.30175781; TotalTime = 1.78603s; TotalTimePerSample = 0.34883ms; SamplesPerSecond = 2866
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.0043954; EvalErrPerSample = 0.31276855; AvgLearningRatePerSample = 0.003125000047; EpochTime=28.984064
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.gpu.txt b/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.gpu.txt
index 17e8f4da6..7513c4dc6 100644
--- a/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.gpu.txt
@@ -3872,4 +3872,4 @@ WARNING: The same matrix with dim [1, 1] has been transferred between different
  Epoch[ 4 of 4]-Minibatch[ 151- 160 of 160]: SamplesSeen = 5120; TrainLossPerSample =  0.97961731; EvalErr[0]PerSample = 0.29921875; TotalTime = 0.15761s; TotalTimePerSample = 0.03078ms; SamplesPerSecond = 32486
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.0073979; EvalErrPerSample = 0.31477052; AvgLearningRatePerSample = 0.003125000047; EpochTime=2.874394
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.windows.cpu.txt
index e02d21617..7d2d64b1a 100644
--- a/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.windows.cpu.txt
@@ -3274,4 +3274,4 @@ Starting minibatch loop.
  Epoch[ 4 of 4]-Minibatch[ 151- 160 of 160]: SamplesSeen = 5120; TrainLossPerSample =  0.96573181; EvalErr[0]PerSample = 0.30156250; TotalTime = 2.77351s; TotalTimePerSample = 0.54170ms; SamplesPerSecond = 1846
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.9992947; EvalErrPerSample = 0.31165773; AvgLearningRatePerSample = 0.003125000047; EpochTime=48.032169
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.windows.gpu.txt
index 656d76c98..5bcdfa56a 100644
--- a/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/baseline.windows.gpu.txt
@@ -3880,4 +3880,4 @@ WARNING: The same matrix with dim [1, 1] has been transferred between different
  Epoch[ 4 of 4]-Minibatch[ 151- 160 of 160]: SamplesSeen = 5120; TrainLossPerSample =  0.96711578; EvalErr[0]PerSample = 0.30175781; TotalTime = 0.49045s; TotalTimePerSample = 0.09579ms; SamplesPerSecond = 10439
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.99611807; EvalErrPerSample = 0.31303713; AvgLearningRatePerSample = 0.003125000047; EpochTime=10.396508
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/testcases.yml b/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/testcases.yml
index fb6cac2af..1e532f940 100644
--- a/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/testcases.yml
+++ b/Tests/EndToEndTests/Speech/DNN/DiscriminativePreTraining/testcases.yml
@@ -19,24 +19,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.25%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.2%}}
 
diff --git a/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.cpu.txt b/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.cpu.txt
index 265ecc04f..8046aaf2f 100644
--- a/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.cpu.txt
@@ -493,7 +493,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.93975925; EvalErr[0]PerSample = 0.53330078; TotalTime = 4.26964s; TotalTimePerSample = 0.41696ms; SamplesPerSecond = 2398
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9483618; EvalErrPerSample = 0.53447266; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=8.545751
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/06 11:38:28
 MPI Rank 1: command line options: 
@@ -959,7 +959,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.93975925; EvalErr[0]PerSample = 0.53330078; TotalTime = 4.25975s; TotalTimePerSample = 0.41599ms; SamplesPerSecond = 2403
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9483618; EvalErrPerSample = 0.53447266; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=8.529597
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/06 11:38:29
 MPI Rank 2: command line options: 
@@ -1425,5 +1425,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.93975925; EvalErr[0]PerSample = 0.53330078; TotalTime = 4.27664s; TotalTimePerSample = 0.41764ms; SamplesPerSecond = 2394
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9483618; EvalErrPerSample = 0.53447266; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=8.524726
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.gpu.txt b/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.gpu.txt
index 58af59e47..ebe66f7ad 100644
--- a/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.gpu.txt
@@ -494,7 +494,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.94366837; EvalErr[0]PerSample = 0.53730469; TotalTime = 0.55136s; TotalTimePerSample = 0.05384ms; SamplesPerSecond = 18572
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9596503; EvalErrPerSample = 0.53989258; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=1.14466
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/06 11:42:51
 MPI Rank 1: command line options: 
@@ -961,7 +961,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.94366837; EvalErr[0]PerSample = 0.53730469; TotalTime = 0.55136s; TotalTimePerSample = 0.05384ms; SamplesPerSecond = 18572
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9596503; EvalErrPerSample = 0.53989258; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=1.144913
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/06 11:42:51
 MPI Rank 2: command line options: 
@@ -1428,5 +1428,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.94366837; EvalErr[0]PerSample = 0.53730469; TotalTime = 0.55136s; TotalTimePerSample = 0.05384ms; SamplesPerSecond = 18572
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9596503; EvalErrPerSample = 0.53989258; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=1.145159
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.windows.cpu.txt
index 45c74451f..88e717fe8 100644
--- a/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.windows.cpu.txt
@@ -500,7 +500,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90745194; EvalErr[0]PerSample = 0.52822266; TotalTime = 5.33597s; TotalTimePerSample = 0.52109ms; SamplesPerSecond = 1919
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9172556; EvalErrPerSample = 0.53793945; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=10.70695
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006095637.485670\Speech\DNN_Parallel1BitQuantization@debug_cpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -976,7 +976,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90745194; EvalErr[0]PerSample = 0.52822266; TotalTime = 5.33601s; TotalTimePerSample = 0.52110ms; SamplesPerSecond = 1919
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9172556; EvalErrPerSample = 0.53793945; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=10.706864
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006095637.485670\Speech\DNN_Parallel1BitQuantization@debug_cpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1452,5 +1452,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90745194; EvalErr[0]PerSample = 0.52822266; TotalTime = 5.33578s; TotalTimePerSample = 0.52107ms; SamplesPerSecond = 1919
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9172556; EvalErrPerSample = 0.53793945; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=10.706562
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.windows.gpu.txt
index d982cd975..02b6bc0ce 100644
--- a/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/Parallel1BitQuantization/baseline.windows.gpu.txt
@@ -501,7 +501,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90162239; EvalErr[0]PerSample = 0.52011719; TotalTime = 1.36215s; TotalTimePerSample = 0.13302ms; SamplesPerSecond = 7517
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.904514; EvalErrPerSample = 0.52109375; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=2.921597
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006095145.884402\Speech\DNN_Parallel1BitQuantization@debug_gpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -978,7 +978,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90162239; EvalErr[0]PerSample = 0.52011719; TotalTime = 1.36219s; TotalTimePerSample = 0.13303ms; SamplesPerSecond = 7517
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.904514; EvalErrPerSample = 0.52109375; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=2.921277
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006095145.884402\Speech\DNN_Parallel1BitQuantization@debug_gpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1455,5 +1455,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90162239; EvalErr[0]PerSample = 0.52011719; TotalTime = 1.36172s; TotalTimePerSample = 0.13298ms; SamplesPerSecond = 7519
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.904514; EvalErrPerSample = 0.52109375; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=2.92187
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.cpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.cpu.txt
index 87cf5264b..0db984cfc 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.cpu.txt
@@ -968,7 +968,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  2.03252134; EvalErr[0]PerSample = 0.54667969; TotalTime = 2.64420s; TotalTimePerSample = 0.25822ms; SamplesPerSecond = 3872
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 2.0474117; EvalErrPerSample = 0.546875; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=5.515902
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -1909,7 +1909,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  2.03252134; EvalErr[0]PerSample = 0.54667969; TotalTime = 2.73810s; TotalTimePerSample = 0.26739ms; SamplesPerSecond = 3739
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 2.0474117; EvalErrPerSample = 0.546875; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=5.515901
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -2850,5 +2850,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  2.03252134; EvalErr[0]PerSample = 0.54667969; TotalTime = 2.76024s; TotalTimePerSample = 0.26955ms; SamplesPerSecond = 3709
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 2.0474117; EvalErrPerSample = 0.546875; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=5.51589
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.gpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.gpu.txt
index 8402d3f32..6e7ffef56 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.gpu.txt
@@ -970,7 +970,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.96679954; EvalErr[0]PerSample = 0.54326172; TotalTime = 0.74867s; TotalTimePerSample = 0.07311ms; SamplesPerSecond = 13677
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9829983; EvalErrPerSample = 0.54199219; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.628572
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -1913,7 +1913,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.96679954; EvalErr[0]PerSample = 0.54326172; TotalTime = 0.73508s; TotalTimePerSample = 0.07179ms; SamplesPerSecond = 13930
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9829983; EvalErrPerSample = 0.54199219; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.632037
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -2856,5 +2856,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.96679954; EvalErr[0]PerSample = 0.54326172; TotalTime = 0.73439s; TotalTimePerSample = 0.07172ms; SamplesPerSecond = 13943
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9829983; EvalErrPerSample = 0.54199219; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.632004
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
index 28e525b45..6d34c8ffd 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
@@ -959,7 +959,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90311195; EvalErr[0]PerSample = 0.51884766; TotalTime = 10.16914s; TotalTimePerSample = 0.99308ms; SamplesPerSecond = 1006
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.916061; EvalErrPerSample = 0.52392578; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=18.802169
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151129190328.957292\Speech\DNN_ParallelBufferedAsyncGradientAggregation@debug_cpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1894,7 +1894,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90311195; EvalErr[0]PerSample = 0.51884766; TotalTime = 10.57049s; TotalTimePerSample = 1.03227ms; SamplesPerSecond = 968
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.916061; EvalErrPerSample = 0.52392578; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=18.802116
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151129190328.957292\Speech\DNN_ParallelBufferedAsyncGradientAggregation@debug_cpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -2829,5 +2829,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90311195; EvalErr[0]PerSample = 0.51884766; TotalTime = 10.02738s; TotalTimePerSample = 0.97924ms; SamplesPerSecond = 1021
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.916061; EvalErrPerSample = 0.52392578; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=18.802161
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
index 0a8246adf..ff073cba5 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
@@ -961,7 +961,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91243517; EvalErr[0]PerSample = 0.52392578; TotalTime = 1.19099s; TotalTimePerSample = 0.11631ms; SamplesPerSecond = 8597
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9294785; EvalErrPerSample = 0.53051758; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.665546
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151129190035.845361\Speech\DNN_ParallelBufferedAsyncGradientAggregation@debug_gpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1898,7 +1898,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91243517; EvalErr[0]PerSample = 0.52392578; TotalTime = 1.18460s; TotalTimePerSample = 0.11568ms; SamplesPerSecond = 8644
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9294785; EvalErrPerSample = 0.53051758; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.665537
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151129190035.845361\Speech\DNN_ParallelBufferedAsyncGradientAggregation@debug_gpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -2835,5 +2835,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91243517; EvalErr[0]PerSample = 0.52392578; TotalTime = 1.20963s; TotalTimePerSample = 0.11813ms; SamplesPerSecond = 8465
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9294785; EvalErrPerSample = 0.53051758; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.665575
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.cpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.cpu.txt
index c3fc02a9c..29ce3bdf7 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.cpu.txt
@@ -684,7 +684,7 @@ MPI Rank 0: CNTKCommandTrainEnd: speechTrain
 MPI Rank 0: 
 MPI Rank 0: Action "train" complete.
 MPI Rank 0: 
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -1301,5 +1301,5 @@ MPI Rank 1: CNTKCommandTrainEnd: speechTrain
 MPI Rank 1: 
 MPI Rank 1: Action "train" complete.
 MPI Rank 1: 
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.gpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.gpu.txt
index 38de79728..59b73ddb2 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.gpu.txt
@@ -685,7 +685,7 @@ MPI Rank 0: CNTKCommandTrainEnd: speechTrain
 MPI Rank 0: 
 MPI Rank 0: Action "train" complete.
 MPI Rank 0: 
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -1303,5 +1303,5 @@ MPI Rank 1: CNTKCommandTrainEnd: speechTrain
 MPI Rank 1: 
 MPI Rank 1: Action "train" complete.
 MPI Rank 1: 
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
\ No newline at end of file
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.windows.cpu.txt
index d7c6d81a9..433f7de38 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.windows.cpu.txt
@@ -622,7 +622,7 @@ MPI Rank 0: Final Results: Minibatch[1-82]: Samples Seen = 83050    CrossEntropy
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.8521576; EvalErrPerSample = 0.50912703
 MPI Rank 0: SGD: Saving checkpoint model 'C:\cygwin64\tmp\cntk-test-20160301170019.423861\Speech\DNN_ParallelCrossValidation@debug_cpu/models/cntkSpeech.dnn'
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160301170019.423861\Speech\DNN_ParallelCrossValidation@debug_cpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1226,5 +1226,5 @@ MPI Rank 1: minibatchiterator: epoch 0: frames [0..83050] (first utterance at fr
 MPI Rank 1: Final Results: Minibatch[1-82]: Samples Seen = 83050    CrossEntropyWithSoftmax: CrossEntropyWithSoftmax/Sample = 1.8521576    Perplexity = 6.3735562    EvalErrorPrediction: ErrorPrediction/Sample = 0.50912703    
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.8521576; EvalErrPerSample = 0.50912703
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.windows.gpu.txt
index c46823ec4..04c9a2610 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelCrossValidation/baseline.windows.gpu.txt
@@ -623,7 +623,7 @@ MPI Rank 0: Final Results: Minibatch[1-82]: Samples Seen = 83050    CrossEntropy
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.8458415; EvalErrPerSample = 0.50965683
 MPI Rank 0: SGD: Saving checkpoint model 'C:\cygwin64\tmp\cntk-test-20160301172412.673018\Speech\DNN_ParallelCrossValidation@release_gpu/models/cntkSpeech.dnn'
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160301172412.673018\Speech\DNN_ParallelCrossValidation@release_gpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1228,5 +1228,5 @@ MPI Rank 1: minibatchiterator: epoch 0: frames [0..83050] (first utterance at fr
 MPI Rank 1: Final Results: Minibatch[1-82]: Samples Seen = 83050    CrossEntropyWithSoftmax: CrossEntropyWithSoftmax/Sample = 1.8458415    Perplexity = 6.3334268    EvalErrorPrediction: ErrorPrediction/Sample = 0.50965683    
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.8458415; EvalErrPerSample = 0.50965683
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.cpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.cpu.txt
index af204e6c1..1adf6bf05 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.cpu.txt
@@ -494,7 +494,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91062447; EvalErr[0]PerSample = 0.52783203; TotalTime = 3.92841s; TotalTimePerSample = 0.38363ms; SamplesPerSecond = 2606
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9150216; EvalErrPerSample = 0.52836914; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=7.722059
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/02 13:34:22
 MPI Rank 1: command line options: 
@@ -958,7 +958,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91062447; EvalErr[0]PerSample = 0.52783203; TotalTime = 3.91134s; TotalTimePerSample = 0.38197ms; SamplesPerSecond = 2618
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9150216; EvalErrPerSample = 0.52836914; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=7.721915
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/02 13:34:22
 MPI Rank 2: command line options: 
@@ -1422,5 +1422,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91062447; EvalErr[0]PerSample = 0.52783203; TotalTime = 3.89837s; TotalTimePerSample = 0.38070ms; SamplesPerSecond = 2626
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9150216; EvalErrPerSample = 0.52836914; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=7.712128
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.gpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.gpu.txt
index 359606b26..0c4334e0c 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.gpu.txt
@@ -606,7 +606,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.86656271; EvalErr[0]PerSample = 0.51748047; TotalTime = 0.34559s; TotalTimePerSample = 0.03375ms; SamplesPerSecond = 29630
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.5184082; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=0.728516
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/24 12:56:12
 MPI Rank 1: command line: 
@@ -1185,7 +1185,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.86656271; EvalErr[0]PerSample = 0.51748047; TotalTime = 0.34559s; TotalTimePerSample = 0.03375ms; SamplesPerSecond = 29630
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.5184082; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=0.728446
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/24 12:56:12
 MPI Rank 2: command line: 
@@ -1764,5 +1764,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.86656271; EvalErr[0]PerSample = 0.51748047; TotalTime = 0.34556s; TotalTimePerSample = 0.03375ms; SamplesPerSecond = 29632
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.5184082; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=0.728405
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.windows.cpu.txt
index c0207fc32..0044448e0 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.windows.cpu.txt
@@ -501,7 +501,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.89380105; EvalErr[0]PerSample = 0.51816406; TotalTime = 5.32609s; TotalTimePerSample = 0.52013ms; SamplesPerSecond = 1922
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8898486; EvalErrPerSample = 0.52172852; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=11.273566
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151002132240.147218\Speech\DNN_ParallelNoQuantization@debug_cpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -975,7 +975,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.89380105; EvalErr[0]PerSample = 0.51816406; TotalTime = 5.32607s; TotalTimePerSample = 0.52012ms; SamplesPerSecond = 1922
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8898486; EvalErrPerSample = 0.52172852; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=11.27354
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151002132240.147218\Speech\DNN_ParallelNoQuantization@debug_cpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1449,5 +1449,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.89380105; EvalErr[0]PerSample = 0.51816406; TotalTime = 5.32607s; TotalTimePerSample = 0.52012ms; SamplesPerSecond = 1922
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8898486; EvalErrPerSample = 0.52172852; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=11.27344
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.windows.gpu.txt
index 2c04b9eea..b22105956 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantization/baseline.windows.gpu.txt
@@ -612,7 +612,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.87358831; EvalErr[0]PerSample = 0.51542969; TotalTime = 0.82689s; TotalTimePerSample = 0.08075ms; SamplesPerSecond = 12383
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8705584; EvalErrPerSample = 0.51860352; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.909274
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151024141411.953694\Speech\DNN_ParallelNoQuantization@debug_gpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1200,7 +1200,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.87358831; EvalErr[0]PerSample = 0.51542969; TotalTime = 0.82899s; TotalTimePerSample = 0.08096ms; SamplesPerSecond = 12352
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8705584; EvalErrPerSample = 0.51860352; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.90997
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151024141411.953694\Speech\DNN_ParallelNoQuantization@debug_gpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1788,5 +1788,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.87358831; EvalErr[0]PerSample = 0.51542969; TotalTime = 0.82561s; TotalTimePerSample = 0.08063ms; SamplesPerSecond = 12402
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8705584; EvalErrPerSample = 0.51860352; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.906446
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.cpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.cpu.txt
index 5c0259ff4..2c131b576 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.cpu.txt
@@ -877,7 +877,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 0: Async gradient aggregation wait time: 0.006704
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9047436; EvalErrPerSample = 0.52290039; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=6.84798
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -1727,7 +1727,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 1: Async gradient aggregation wait time: 1.3e-05
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9047436; EvalErrPerSample = 0.52290039; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=6.8934
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -2577,5 +2577,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 2: Async gradient aggregation wait time: 0.004703
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9047436; EvalErrPerSample = 0.52290039; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=6.84798
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.gpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.gpu.txt
index d9019ff9c..fb0d7fb87 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.gpu.txt
@@ -883,7 +883,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 0: Async gradient aggregation wait time: 0.004648
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8924891; EvalErrPerSample = 0.51933594; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.10761
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -1739,7 +1739,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 1: Async gradient aggregation wait time: 0.004311
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8924891; EvalErrPerSample = 0.51933594; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.10764
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -2595,5 +2595,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 2: Async gradient aggregation wait time: 0.005346
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8924891; EvalErrPerSample = 0.51933594; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.10805
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
index 56262d3f4..46839332c 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
@@ -864,7 +864,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 0: Async gradient aggregation wait time: 9e-006
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8894112; EvalErrPerSample = 0.51376953; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=22.561
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160106231018.691457\Speech\DNN_ParallelNoQuantizationBufferedAsyncGradientAggregation@debug_cpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1699,7 +1699,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 1: Async gradient aggregation wait time: 8e-006
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8894112; EvalErrPerSample = 0.51376953; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=22.7997
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160106231018.691457\Speech\DNN_ParallelNoQuantizationBufferedAsyncGradientAggregation@debug_cpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -2534,5 +2534,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 2: Async gradient aggregation wait time: 0.00373
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8894112; EvalErrPerSample = 0.51376953; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=22.5056
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
index 6618371ea..fb027ae62 100644
--- a/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
@@ -879,7 +879,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 0: Async gradient aggregation wait time: 0.002651
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.882856; EvalErrPerSample = 0.51445312; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.10527
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160106230707.566663\Speech\DNN_ParallelNoQuantizationBufferedAsyncGradientAggregation@debug_gpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1729,7 +1729,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 1: Async gradient aggregation wait time: 0.003634
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.882856; EvalErrPerSample = 0.51445312; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.10546
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160106230707.566663\Speech\DNN_ParallelNoQuantizationBufferedAsyncGradientAggregation@debug_gpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -2584,5 +2584,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 2: Async gradient aggregation wait time: 0.001318
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.882856; EvalErrPerSample = 0.51445312; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.10527
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/DNN/SequenceTraining/baseline.gpu.txt b/Tests/EndToEndTests/Speech/DNN/SequenceTraining/baseline.gpu.txt
index 0d3cbde62..a30ded54e 100644
--- a/Tests/EndToEndTests/Speech/DNN/SequenceTraining/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/SequenceTraining/baseline.gpu.txt
@@ -9586,4 +9586,4 @@ parallelforwardbackwardlattice: 25 launches for forward, 25 launches for backwar
 dengamma value 1.062592
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 0.082648888; EvalErrPerSample = 0.32607052; AvgLearningRatePerSample = 1.999999995e-06; EpochTime=15.94079
 CNTKCommandTrainEnd: sequenceTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/DNN/SequenceTraining/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/DNN/SequenceTraining/baseline.windows.gpu.txt
index af3d47ba6..89ff45731 100644
--- a/Tests/EndToEndTests/Speech/DNN/SequenceTraining/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/SequenceTraining/baseline.windows.gpu.txt
@@ -9598,4 +9598,4 @@ dengamma value 1.142857
  Epoch[ 3 of 3]-Minibatch[ 121- 130 of 8192]: SamplesSeen = 6564; TrainLossPerSample =  0.08458945; EvalErr[0]PerSample = 0.30042657; TotalTime = 2.28939s; TotalTimePerSample = 0.34878ms; SamplesPerSecond = 2867
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 0.083290465; EvalErrPerSample = 0.31113231; AvgLearningRatePerSample = 1.999999995e-006; EpochTime=30.742683
 CNTKCommandTrainEnd: sequenceTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/DNN/SequenceTraining/testcases.yml b/Tests/EndToEndTests/Speech/DNN/SequenceTraining/testcases.yml
index 5b20cd5e2..d0ddb5dec 100644
--- a/Tests/EndToEndTests/Speech/DNN/SequenceTraining/testcases.yml
+++ b/Tests/EndToEndTests/Speech/DNN/SequenceTraining/testcases.yml
@@ -9,24 +9,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.25%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.2%}}
 
diff --git a/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.cpu.txt b/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.cpu.txt
index 8e6032928..4fb9661c3 100644
--- a/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.cpu.txt
@@ -575,4 +575,4 @@ Total Samples Evaluated = 3250
 
 Action "write" complete.
 
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.gpu.txt b/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.gpu.txt
index c891f1383..cb9b5dcf6 100644
--- a/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.gpu.txt
@@ -577,4 +577,4 @@ Total Samples Evaluated = 3250
 
 Action "write" complete.
 
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.windows.cpu.txt
index 77614bfaf..ada98134d 100644
--- a/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.windows.cpu.txt
@@ -663,4 +663,4 @@ Total Samples Evaluated = 3250
 
 Action "write" complete.
 
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.windows.gpu.txt
index 6922ba02d..1ac38654d 100644
--- a/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/DNN/WriteCommand/baseline.windows.gpu.txt
@@ -665,4 +665,4 @@ Total Samples Evaluated = 3250
 
 Action "write" complete.
 
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/DNN/WriteCommand/testcases.yml b/Tests/EndToEndTests/Speech/DNN/WriteCommand/testcases.yml
index 4af43587a..d42e2d2df 100644
--- a/Tests/EndToEndTests/Speech/DNN/WriteCommand/testcases.yml
+++ b/Tests/EndToEndTests/Speech/DNN/WriteCommand/testcases.yml
@@ -8,24 +8,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.1%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErr[0]PerSample = {{float,tolerance=.1%}}
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.cpu.txt
index 82fdb6138..dfea1ce77 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.cpu.txt
@@ -3265,4 +3265,4 @@ Starting minibatch loop.
  Epoch[ 4 of 4]-Minibatch[ 151- 160 of 160]: SamplesSeen = 5120; TrainLossPerSample =  0.97683563; EvalErr[0]PerSample = 0.30175781; TotalTime = 1.78603s; TotalTimePerSample = 0.34883ms; SamplesPerSecond = 2866
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.0043954; EvalErrPerSample = 0.31276855; AvgLearningRatePerSample = 0.003125000047; EpochTime=28.984064
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.gpu.txt
index 17e8f4da6..7513c4dc6 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.gpu.txt
@@ -3872,4 +3872,4 @@ WARNING: The same matrix with dim [1, 1] has been transferred between different
  Epoch[ 4 of 4]-Minibatch[ 151- 160 of 160]: SamplesSeen = 5120; TrainLossPerSample =  0.97961731; EvalErr[0]PerSample = 0.29921875; TotalTime = 0.15761s; TotalTimePerSample = 0.03078ms; SamplesPerSecond = 32486
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.0073979; EvalErrPerSample = 0.31477052; AvgLearningRatePerSample = 0.003125000047; EpochTime=2.874394
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.windows.cpu.txt
index e02d21617..7d2d64b1a 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.windows.cpu.txt
@@ -3274,4 +3274,4 @@ Starting minibatch loop.
  Epoch[ 4 of 4]-Minibatch[ 151- 160 of 160]: SamplesSeen = 5120; TrainLossPerSample =  0.96573181; EvalErr[0]PerSample = 0.30156250; TotalTime = 2.77351s; TotalTimePerSample = 0.54170ms; SamplesPerSecond = 1846
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.9992947; EvalErrPerSample = 0.31165773; AvgLearningRatePerSample = 0.003125000047; EpochTime=48.032169
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.windows.gpu.txt
index 656d76c98..5bcdfa56a 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/baseline.windows.gpu.txt
@@ -3880,4 +3880,4 @@ WARNING: The same matrix with dim [1, 1] has been transferred between different
  Epoch[ 4 of 4]-Minibatch[ 151- 160 of 160]: SamplesSeen = 5120; TrainLossPerSample =  0.96711578; EvalErr[0]PerSample = 0.30175781; TotalTime = 0.49045s; TotalTimePerSample = 0.09579ms; SamplesPerSecond = 10439
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 0.99611807; EvalErrPerSample = 0.31303713; AvgLearningRatePerSample = 0.003125000047; EpochTime=10.396508
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/run-test b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/run-test
index 77d70c60d..48ebc1ec5 100755
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/run-test
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/run-test
@@ -5,11 +5,5 @@
 OriginalTestDir=../../../DNN/DiscriminativePreTraining
 ConfigDir=$TEST_DIR/$OriginalTestDir
 
-(cd $TEST_DIR/$OriginalTestDir && md5sum baseline*) | (cd $TEST_DIR && md5sum --status -c -)
-if [ $? != 0 ]; then
-  echo Error: Baselines must match original test.
-  exit 1
-fi
-
 # cntkrun <CNTK config file name> <additional CNTK args>
 cntkrun cntk_dpt.cntk 'reader=[readerType=ExperimentalHTKMLFReader] reader=[prefetch=true]' || exit $?
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/testcases.yml b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/testcases.yml
index d62ac95d2..b07812a1a 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/testcases.yml
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/DiscriminativePreTraining/testcases.yml
@@ -19,24 +19,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.25%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.2%}}
 
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.cpu.txt
index 265ecc04f..8046aaf2f 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.cpu.txt
@@ -493,7 +493,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.93975925; EvalErr[0]PerSample = 0.53330078; TotalTime = 4.26964s; TotalTimePerSample = 0.41696ms; SamplesPerSecond = 2398
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9483618; EvalErrPerSample = 0.53447266; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=8.545751
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/06 11:38:28
 MPI Rank 1: command line options: 
@@ -959,7 +959,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.93975925; EvalErr[0]PerSample = 0.53330078; TotalTime = 4.25975s; TotalTimePerSample = 0.41599ms; SamplesPerSecond = 2403
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9483618; EvalErrPerSample = 0.53447266; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=8.529597
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/06 11:38:29
 MPI Rank 2: command line options: 
@@ -1425,5 +1425,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.93975925; EvalErr[0]PerSample = 0.53330078; TotalTime = 4.27664s; TotalTimePerSample = 0.41764ms; SamplesPerSecond = 2394
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9483618; EvalErrPerSample = 0.53447266; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=8.524726
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.gpu.txt
index 58af59e47..ebe66f7ad 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.gpu.txt
@@ -494,7 +494,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.94366837; EvalErr[0]PerSample = 0.53730469; TotalTime = 0.55136s; TotalTimePerSample = 0.05384ms; SamplesPerSecond = 18572
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9596503; EvalErrPerSample = 0.53989258; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=1.14466
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/06 11:42:51
 MPI Rank 1: command line options: 
@@ -961,7 +961,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.94366837; EvalErr[0]PerSample = 0.53730469; TotalTime = 0.55136s; TotalTimePerSample = 0.05384ms; SamplesPerSecond = 18572
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9596503; EvalErrPerSample = 0.53989258; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=1.144913
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/06 11:42:51
 MPI Rank 2: command line options: 
@@ -1428,5 +1428,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.94366837; EvalErr[0]PerSample = 0.53730469; TotalTime = 0.55136s; TotalTimePerSample = 0.05384ms; SamplesPerSecond = 18572
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9596503; EvalErrPerSample = 0.53989258; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=1.145159
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.windows.cpu.txt
index 45c74451f..88e717fe8 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.windows.cpu.txt
@@ -500,7 +500,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90745194; EvalErr[0]PerSample = 0.52822266; TotalTime = 5.33597s; TotalTimePerSample = 0.52109ms; SamplesPerSecond = 1919
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9172556; EvalErrPerSample = 0.53793945; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=10.70695
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006095637.485670\Speech\DNN_Parallel1BitQuantization@debug_cpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -976,7 +976,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90745194; EvalErr[0]PerSample = 0.52822266; TotalTime = 5.33601s; TotalTimePerSample = 0.52110ms; SamplesPerSecond = 1919
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9172556; EvalErrPerSample = 0.53793945; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=10.706864
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006095637.485670\Speech\DNN_Parallel1BitQuantization@debug_cpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1452,5 +1452,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90745194; EvalErr[0]PerSample = 0.52822266; TotalTime = 5.33578s; TotalTimePerSample = 0.52107ms; SamplesPerSecond = 1919
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9172556; EvalErrPerSample = 0.53793945; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=10.706562
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.windows.gpu.txt
index d982cd975..02b6bc0ce 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/baseline.windows.gpu.txt
@@ -501,7 +501,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90162239; EvalErr[0]PerSample = 0.52011719; TotalTime = 1.36215s; TotalTimePerSample = 0.13302ms; SamplesPerSecond = 7517
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.904514; EvalErrPerSample = 0.52109375; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=2.921597
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006095145.884402\Speech\DNN_Parallel1BitQuantization@debug_gpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -978,7 +978,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90162239; EvalErr[0]PerSample = 0.52011719; TotalTime = 1.36219s; TotalTimePerSample = 0.13303ms; SamplesPerSecond = 7517
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.904514; EvalErrPerSample = 0.52109375; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=2.921277
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151006095145.884402\Speech\DNN_Parallel1BitQuantization@debug_gpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1455,5 +1455,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90162239; EvalErr[0]PerSample = 0.52011719; TotalTime = 1.36172s; TotalTimePerSample = 0.13298ms; SamplesPerSecond = 7519
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.904514; EvalErrPerSample = 0.52109375; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=2.92187
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/run-test b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/run-test
index d10215d15..86de681dd 100755
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/run-test
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/Parallel1BitQuantization/run-test
@@ -8,12 +8,6 @@ LogFileName=stderr
 Instances=3
 NumCPUThreads=$(threadsPerInstance $Instances)
 
-(cd $TEST_DIR/$OriginalTestDir && md5sum baseline*) | (cd $TEST_DIR && md5sum --status -c -)
-if [ $? != 0 ]; then
-  echo Error: Baselines must match original test.
-  exit 1
-fi
-
 # cntkmpirun <MPI args> <CNTK config file name> <additional CNTK args>
 cntkmpirun "-n $Instances" cntk.cntk "speechTrain=[reader=[readerType=ExperimentalHTKMLFReader]] speechTrain=[reader=[prefetch=true]] numCPUThreads=$NumCPUThreads precision=double speechTrain=[SGD=[ParallelTrain=[DataParallelSGD=[gradientBits=1]]]] speechTrain=[SGD=[ParallelTrain=[parallelizationStartEpoch=2]]]"
 ExitCode=$?
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.cpu.txt
index 87cf5264b..0db984cfc 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.cpu.txt
@@ -968,7 +968,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  2.03252134; EvalErr[0]PerSample = 0.54667969; TotalTime = 2.64420s; TotalTimePerSample = 0.25822ms; SamplesPerSecond = 3872
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 2.0474117; EvalErrPerSample = 0.546875; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=5.515902
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -1909,7 +1909,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  2.03252134; EvalErr[0]PerSample = 0.54667969; TotalTime = 2.73810s; TotalTimePerSample = 0.26739ms; SamplesPerSecond = 3739
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 2.0474117; EvalErrPerSample = 0.546875; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=5.515901
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -2850,5 +2850,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  2.03252134; EvalErr[0]PerSample = 0.54667969; TotalTime = 2.76024s; TotalTimePerSample = 0.26955ms; SamplesPerSecond = 3709
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 2.0474117; EvalErrPerSample = 0.546875; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=5.51589
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.gpu.txt
index 8402d3f32..6e7ffef56 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.gpu.txt
@@ -970,7 +970,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.96679954; EvalErr[0]PerSample = 0.54326172; TotalTime = 0.74867s; TotalTimePerSample = 0.07311ms; SamplesPerSecond = 13677
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9829983; EvalErrPerSample = 0.54199219; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.628572
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -1913,7 +1913,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.96679954; EvalErr[0]PerSample = 0.54326172; TotalTime = 0.73508s; TotalTimePerSample = 0.07179ms; SamplesPerSecond = 13930
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9829983; EvalErrPerSample = 0.54199219; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.632037
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -2856,5 +2856,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.96679954; EvalErr[0]PerSample = 0.54326172; TotalTime = 0.73439s; TotalTimePerSample = 0.07172ms; SamplesPerSecond = 13943
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9829983; EvalErrPerSample = 0.54199219; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.632004
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
index 28e525b45..6d34c8ffd 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
@@ -959,7 +959,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90311195; EvalErr[0]PerSample = 0.51884766; TotalTime = 10.16914s; TotalTimePerSample = 0.99308ms; SamplesPerSecond = 1006
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.916061; EvalErrPerSample = 0.52392578; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=18.802169
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151129190328.957292\Speech\DNN_ParallelBufferedAsyncGradientAggregation@debug_cpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1894,7 +1894,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90311195; EvalErr[0]PerSample = 0.51884766; TotalTime = 10.57049s; TotalTimePerSample = 1.03227ms; SamplesPerSecond = 968
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.916061; EvalErrPerSample = 0.52392578; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=18.802116
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151129190328.957292\Speech\DNN_ParallelBufferedAsyncGradientAggregation@debug_cpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -2829,5 +2829,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.90311195; EvalErr[0]PerSample = 0.51884766; TotalTime = 10.02738s; TotalTimePerSample = 0.97924ms; SamplesPerSecond = 1021
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.916061; EvalErrPerSample = 0.52392578; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=18.802161
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
index 0a8246adf..ff073cba5 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
@@ -961,7 +961,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91243517; EvalErr[0]PerSample = 0.52392578; TotalTime = 1.19099s; TotalTimePerSample = 0.11631ms; SamplesPerSecond = 8597
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9294785; EvalErrPerSample = 0.53051758; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.665546
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151129190035.845361\Speech\DNN_ParallelBufferedAsyncGradientAggregation@debug_gpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1898,7 +1898,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91243517; EvalErr[0]PerSample = 0.52392578; TotalTime = 1.18460s; TotalTimePerSample = 0.11568ms; SamplesPerSecond = 8644
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9294785; EvalErrPerSample = 0.53051758; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.665537
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151129190035.845361\Speech\DNN_ParallelBufferedAsyncGradientAggregation@debug_gpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -2835,5 +2835,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[   1-  10 of 20]: SamplesSeen = 9216; Trai
 MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91243517; EvalErr[0]PerSample = 0.52392578; TotalTime = 1.20963s; TotalTimePerSample = 0.11813ms; SamplesPerSecond = 8465
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9294785; EvalErrPerSample = 0.53051758; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.665575
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/run-test b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/run-test
index 8396be527..a981eabf3 100755
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/run-test
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelBufferedAsyncGradientAggregation/run-test
@@ -8,12 +8,6 @@ LogFileName=stderr
 Instances=3
 NumCPUThreads=$(threadsPerInstance $Instances)
 
-(cd $TEST_DIR/$OriginalTestDir && md5sum baseline*) | (cd $TEST_DIR && md5sum --status -c -)
-if [ $? != 0 ]; then
-  echo Error: Baselines must match original test.
-  exit 1
-fi
-
 # cntkmpirun <MPI args> <CNTK config file name> <additional CNTK args>
 cntkmpirun "-n $Instances" cntk.cntk "speechTrain=[reader=[readerType=ExperimentalHTKMLFReader]] speechTrain=[reader=[prefetch=true]] numCPUThreads=$NumCPUThreads precision=double speechTrain=[SGD=[ParallelTrain=[DataParallelSGD=[gradientBits=1]]]] speechTrain=[SGD=[ParallelTrain=[DataParallelSGD=[useBufferedAsyncGradientAggregation=true]]]] speechTrain=[SGD=[ParallelTrain=[parallelizationStartEpoch=2]]] speechTrain=[SGD=[maxEpochs=4]] speechTrain=[SGD=[ParallelTrain=[syncPerfStats=5]]]"
 ExitCode=$?
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.cpu.txt
index af204e6c1..1adf6bf05 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.cpu.txt
@@ -494,7 +494,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91062447; EvalErr[0]PerSample = 0.52783203; TotalTime = 3.92841s; TotalTimePerSample = 0.38363ms; SamplesPerSecond = 2606
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9150216; EvalErrPerSample = 0.52836914; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=7.722059
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/02 13:34:22
 MPI Rank 1: command line options: 
@@ -958,7 +958,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91062447; EvalErr[0]PerSample = 0.52783203; TotalTime = 3.91134s; TotalTimePerSample = 0.38197ms; SamplesPerSecond = 2618
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9150216; EvalErrPerSample = 0.52836914; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=7.721915
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/02 13:34:22
 MPI Rank 2: command line options: 
@@ -1422,5 +1422,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91062447; EvalErr[0]PerSample = 0.52783203; TotalTime = 3.89837s; TotalTimePerSample = 0.38070ms; SamplesPerSecond = 2626
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9150216; EvalErrPerSample = 0.52836914; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=7.712128
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.gpu.txt
index 359606b26..0c4334e0c 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.gpu.txt
@@ -606,7 +606,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.86656271; EvalErr[0]PerSample = 0.51748047; TotalTime = 0.34559s; TotalTimePerSample = 0.03375ms; SamplesPerSecond = 29630
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.5184082; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=0.728516
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: running on localhost at 2015/10/24 12:56:12
 MPI Rank 1: command line: 
@@ -1185,7 +1185,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.86656271; EvalErr[0]PerSample = 0.51748047; TotalTime = 0.34559s; TotalTimePerSample = 0.03375ms; SamplesPerSecond = 29630
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.5184082; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=0.728446
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: running on localhost at 2015/10/24 12:56:12
 MPI Rank 2: command line: 
@@ -1764,5 +1764,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.86656271; EvalErr[0]PerSample = 0.51748047; TotalTime = 0.34556s; TotalTimePerSample = 0.03375ms; SamplesPerSecond = 29632
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.5184082; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=0.728405
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.windows.cpu.txt
index c0207fc32..0044448e0 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.windows.cpu.txt
@@ -501,7 +501,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.89380105; EvalErr[0]PerSample = 0.51816406; TotalTime = 5.32609s; TotalTimePerSample = 0.52013ms; SamplesPerSecond = 1922
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8898486; EvalErrPerSample = 0.52172852; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=11.273566
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151002132240.147218\Speech\DNN_ParallelNoQuantization@debug_cpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -975,7 +975,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.89380105; EvalErr[0]PerSample = 0.51816406; TotalTime = 5.32607s; TotalTimePerSample = 0.52012ms; SamplesPerSecond = 1922
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8898486; EvalErrPerSample = 0.52172852; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=11.27354
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151002132240.147218\Speech\DNN_ParallelNoQuantization@debug_cpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1449,5 +1449,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.89380105; EvalErr[0]PerSample = 0.51816406; TotalTime = 5.32607s; TotalTimePerSample = 0.52012ms; SamplesPerSecond = 1922
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8898486; EvalErrPerSample = 0.52172852; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=11.27344
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.windows.gpu.txt
index 2c04b9eea..b22105956 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/baseline.windows.gpu.txt
@@ -612,7 +612,7 @@ MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 0:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.87358831; EvalErr[0]PerSample = 0.51542969; TotalTime = 0.82689s; TotalTimePerSample = 0.08075ms; SamplesPerSecond = 12383
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8705584; EvalErrPerSample = 0.51860352; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.909274
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151024141411.953694\Speech\DNN_ParallelNoQuantization@debug_gpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1200,7 +1200,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.87358831; EvalErr[0]PerSample = 0.51542969; TotalTime = 0.82899s; TotalTimePerSample = 0.08096ms; SamplesPerSecond = 12352
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8705584; EvalErrPerSample = 0.51860352; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.90997
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20151024141411.953694\Speech\DNN_ParallelNoQuantization@debug_gpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1788,5 +1788,5 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; Tra
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.87358831; EvalErr[0]PerSample = 0.51542969; TotalTime = 0.82561s; TotalTimePerSample = 0.08063ms; SamplesPerSecond = 12402
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8705584; EvalErrPerSample = 0.51860352; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.906446
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/run-test b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/run-test
index 23ee8cdda..de5b2af69 100755
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/run-test
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantization/run-test
@@ -8,12 +8,6 @@ LogFileName=stderr
 Instances=3
 NumCPUThreads=$(threadsPerInstance $Instances)
 
-(cd $TEST_DIR/$OriginalTestDir && md5sum baseline*) | (cd $TEST_DIR && md5sum --status -c -)
-if [ $? != 0 ]; then
-  echo Error: Baselines must match original test.
-  exit 1
-fi
-
 # cntkmpirun <MPI args> <CNTK config file name> <additional CNTK args>
 cntkmpirun "-n $Instances" cntk.cntk "speechTrain=[reader=[readerType=ExperimentalHTKMLFReader]] speechTrain=[reader=[prefetch=true]] numCPUThreads=$NumCPUThreads"
 ExitCode=$?
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.cpu.txt
index 5c0259ff4..2c131b576 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.cpu.txt
@@ -877,7 +877,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 0: Async gradient aggregation wait time: 0.006704
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9047436; EvalErrPerSample = 0.52290039; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=6.84798
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -1727,7 +1727,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 1: Async gradient aggregation wait time: 1.3e-05
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9047436; EvalErrPerSample = 0.52290039; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=6.8934
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -2577,5 +2577,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 2: Async gradient aggregation wait time: 0.004703
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.9047436; EvalErrPerSample = 0.52290039; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=6.84798
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.gpu.txt
index d9019ff9c..fb0d7fb87 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.gpu.txt
@@ -883,7 +883,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 0: Async gradient aggregation wait time: 0.004648
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8924891; EvalErrPerSample = 0.51933594; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.10761
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -1739,7 +1739,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 1: Async gradient aggregation wait time: 0.004311
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8924891; EvalErrPerSample = 0.51933594; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.10764
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -2595,5 +2595,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 2: Async gradient aggregation wait time: 0.005346
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8924891; EvalErrPerSample = 0.51933594; AvgLearningRatePerSample = 9.7656251e-05; EpochTime=1.10805
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
index 56262d3f4..46839332c 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.cpu.txt
@@ -864,7 +864,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 0: Async gradient aggregation wait time: 9e-006
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8894112; EvalErrPerSample = 0.51376953; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=22.561
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160106231018.691457\Speech\DNN_ParallelNoQuantizationBufferedAsyncGradientAggregation@debug_cpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1699,7 +1699,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 1: Async gradient aggregation wait time: 8e-006
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8894112; EvalErrPerSample = 0.51376953; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=22.7997
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160106231018.691457\Speech\DNN_ParallelNoQuantizationBufferedAsyncGradientAggregation@debug_cpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -2534,5 +2534,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 2: Async gradient aggregation wait time: 0.00373
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.8894112; EvalErrPerSample = 0.51376953; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=22.5056
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
index 6618371ea..fb027ae62 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/baseline.windows.gpu.txt
@@ -879,7 +879,7 @@ MPI Rank 0:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 0: Async gradient aggregation wait time: 0.002651
 MPI Rank 0: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.882856; EvalErrPerSample = 0.51445312; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.10527
 MPI Rank 0: CNTKCommandTrainEnd: speechTrain
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160106230707.566663\Speech\DNN_ParallelNoQuantizationBufferedAsyncGradientAggregation@debug_gpu/stderr_speechTrain.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -1729,7 +1729,7 @@ MPI Rank 1:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 1: Async gradient aggregation wait time: 0.003634
 MPI Rank 1: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.882856; EvalErrPerSample = 0.51445312; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.10546
 MPI Rank 1: CNTKCommandTrainEnd: speechTrain
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160106230707.566663\Speech\DNN_ParallelNoQuantizationBufferedAsyncGradientAggregation@debug_gpu/stderr_speechTrain.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -2584,5 +2584,5 @@ MPI Rank 2:  Epoch[ 4 of 4]-Minibatch[  11-  20, 100.00%]: SamplesSeen = 10240;
 MPI Rank 2: Async gradient aggregation wait time: 0.001318
 MPI Rank 2: Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 1.882856; EvalErrPerSample = 0.51445312; AvgLearningRatePerSample = 9.7656251e-005; EpochTime=2.10527
 MPI Rank 2: CNTKCommandTrainEnd: speechTrain
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/run-test b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/run-test
index ec4faa876..3d7d8c5d4 100755
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/run-test
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/DNN/ParallelNoQuantizationBufferedAsyncGradientAggregation/run-test
@@ -8,12 +8,6 @@ LogFileName=stderr
 Instances=3
 NumCPUThreads=$(threadsPerInstance $Instances)
 
-(cd $TEST_DIR/$OriginalTestDir && md5sum baseline*) | (cd $TEST_DIR && md5sum --status -c -)
-if [ $? != 0 ]; then
-  echo Error: Baselines must match original test.
-  exit 1
-fi
-
 # cntkmpirun <MPI args> <CNTK config file name> <additional CNTK args>
 cntkmpirun "-n $Instances" cntk.cntk "speechTrain=[reader=[readerType=ExperimentalHTKMLFReader]] speechTrain=[reader=[prefetch=true]] numCPUThreads=$NumCPUThreads precision=double speechTrain=[SGD=[ParallelTrain=[DataParallelSGD=[gradientBits=64]]]] speechTrain=[SGD=[ParallelTrain=[DataParallelSGD=[useBufferedAsyncGradientAggregation=true]]]] speechTrain=[SGD=[ParallelTrain=[parallelizationStartEpoch=2]]] speechTrain=[SGD=[maxEpochs=4]] speechTrain=[SGD=[ParallelTrain=[syncPerfStats=5]]]"
 ExitCode=$?
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.cpu.txt
index 74268f0c3..1ffd26847 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.cpu.txt
@@ -340,7 +340,7 @@ Starting minibatch loop.
  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91941869; EvalErr[0]PerSample = 0.52890623; TotalTime = 1.86784s; TotalTimePerSample = 0.18241ms; SamplesPerSecond = 5482
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91062427; EvalErr[0]PerSample = 0.52783203; TotalTime = 1.84987s; TotalTimePerSample = 0.18065ms; SamplesPerSecond = 5535
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9150215; EvalErrPerSample = 0.52836913; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=3.736133
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 running on localhost at 2015/08/25 20:26:31
@@ -710,4 +710,4 @@ Validating --> EvalErrorPrediction = ErrorPrediction(labels[132, 1024], HLast[13
  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91941869; EvalErr[0]PerSample = 0.52890623; TotalTime = 1.88723s; TotalTimePerSample = 0.18430ms; SamplesPerSecond = 5425
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91062427; EvalErr[0]PerSample = 0.52783203; TotalTime = 1.84469s; TotalTimePerSample = 0.18015ms; SamplesPerSecond = 5551
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9150215; EvalErrPerSample = 0.52836913; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=5.315324
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.gpu.txt
index d71d3eac2..b2076d6f4 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.gpu.txt
@@ -553,7 +553,7 @@ Starting minibatch loop.
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.86656265; EvalErr[0]PerSample = 0.51748047; TotalTime = 0.21814s; TotalTimePerSample = 0.02130ms; SamplesPerSecond = 46943
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.51840824; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=0.493964
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 === Running /home/mluser/src/cplx_master/build/debug/bin/cntk configFile=/home/mluser/src/cplx_master/Tests/Speech/QuickE2E/cntk.config RunDir=/tmp/cntk-test-20151024124900.548963/Speech_QuickE2E@debug_gpu DataDir=/home/mluser/src/cplx_master/Tests/Speech/Data ConfigDir=/home/mluser/src/cplx_master/Tests/Speech/QuickE2E DeviceId=0
@@ -1331,4 +1331,4 @@ EnforceOneGPUOnly: WARNING: Ignored attempt to change GPU choice from 0 now 1. T
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.86656265; EvalErr[0]PerSample = 0.51748047; TotalTime = 0.21717s; TotalTimePerSample = 0.02121ms; SamplesPerSecond = 47152
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.51840824; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=1.439589
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.windows.cpu.txt
index ff037b498..f899e6ae1 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.windows.cpu.txt
@@ -349,7 +349,7 @@ Starting minibatch loop, distributed reading is: DISABLED
  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.88589633; EvalErr[0]PerSample = 0.52529299; TotalTime = 0.59385s; TotalTimePerSample = 0.05799ms; SamplesPerSecond = 17243
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.89380074; EvalErr[0]PerSample = 0.51816404; TotalTime = 0.51284s; TotalTimePerSample = 0.05008ms; SamplesPerSecond = 19967
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8898486; EvalErrPerSample = 0.52172852; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.109687
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 -------------------------------------------------------------------
@@ -728,4 +728,4 @@ Validating --> EvalErrorPrediction = ErrorPrediction(labels[132, 1024], HLast[13
  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.88589633; EvalErr[0]PerSample = 0.52529299; TotalTime = 0.86012s; TotalTimePerSample = 0.08400ms; SamplesPerSecond = 11905
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.89380074; EvalErr[0]PerSample = 0.51816404; TotalTime = 0.60616s; TotalTimePerSample = 0.05919ms; SamplesPerSecond = 16893
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8898486; EvalErrPerSample = 0.52172852; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=2.711551
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.windows.gpu.txt
index 0e9622b78..759d87b54 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/baseline.windows.gpu.txt
@@ -561,7 +561,7 @@ Starting minibatch loop.
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.87358818; EvalErr[0]PerSample = 0.51542969; TotalTime = 0.60551s; TotalTimePerSample = 0.05913ms; SamplesPerSecond = 16911
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8705584; EvalErrPerSample = 0.5186035; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.428405
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 === Running /cygdrive/e/NetScale/CNTK/git_repos/cplx_master2/x64/debug/cntk.exe configFile=E:\NetScale\CNTK\git_repos\cplx_master2\Tests\Speech\QuickE2E/cntk.config RunDir=C:\cygwin64\tmp\cntk-test-20151024140721.927553\Speech_QuickE2E@debug_gpu DataDir=E:\NetScale\CNTK\git_repos\cplx_master2\Tests\Speech\Data ConfigDir=E:\NetScale\CNTK\git_repos\cplx_master2\Tests\Speech\QuickE2E DeviceId=0
@@ -1347,4 +1347,4 @@ Starting minibatch loop.
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.87358818; EvalErr[0]PerSample = 0.51542969; TotalTime = 0.86938s; TotalTimePerSample = 0.08490ms; SamplesPerSecond = 11778
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8705584; EvalErrPerSample = 0.5186035; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=6.283729
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/run-test b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/run-test
index a59930676..5b2f6432b 100755
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/run-test
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/run-test
@@ -5,12 +5,6 @@
 OriginalTestDir=../../QuickE2E
 ConfigDir=$TEST_DIR/$OriginalTestDir
 
-(cd $TEST_DIR/$OriginalTestDir && md5sum baseline*) | (cd $TEST_DIR && md5sum --status -c -)
-if [ $? != 0 ]; then
-  echo Error: Baselines must match original test.
-  exit 1
-fi
-
 # cntkrun <CNTK config file name> <additional CNTK args>
 DeleteModelsAfterTest=0
 cntkrun cntk.cntk 'speechTrain=[reader=[readerType=ExperimentalHTKMLFReader]]' || exit $?
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/testcases.yml b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/testcases.yml
index b95982f44..46ddcfdfc 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/testcases.yml
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/QuickE2E/testcases.yml
@@ -8,24 +8,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.1%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErr[0]PerSample = {{float,tolerance=.1%}}
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.cpu.txt
index 5c035c099..0dc1812cc 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.cpu.txt
@@ -2064,4 +2064,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.85902176; EvalErr[0]PerSample = 0.51396484; TotalTime = 1.48188s; TotalTimePerSample = 0.14471ms; SamplesPerSecond = 6910
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 1.8499603; EvalErrPerSample = 0.5126465; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=3.049746
 CNTKCommandTrainEnd: SVDTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.gpu.txt
index 779884461..e5315f319 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.gpu.txt
@@ -2067,4 +2067,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.81663570; EvalErr[0]PerSample = 0.50869141; TotalTime = 0.23803s; TotalTimePerSample = 0.02325ms; SamplesPerSecond = 43019
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 1.8090889; EvalErrPerSample = 0.504834; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=0.528531
 CNTKCommandTrainEnd: SVDTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.windows.cpu.txt
index 3dc50ea83..d6751a433 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.windows.cpu.txt
@@ -2073,4 +2073,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.80246696; EvalErr[0]PerSample = 0.50654297; TotalTime = 3.29585s; TotalTimePerSample = 0.32186ms; SamplesPerSecond = 3106
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 1.8185977; EvalErrPerSample = 0.50717777; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=7.151551
 CNTKCommandTrainEnd: SVDTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.windows.gpu.txt
index 31c006e0c..7a0b10c09 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/baseline.windows.gpu.txt
@@ -2075,4 +2075,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.78207207; EvalErr[0]PerSample = 0.50322266; TotalTime = 0.63083s; TotalTimePerSample = 0.06160ms; SamplesPerSecond = 16232
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 1.7974522; EvalErrPerSample = 0.50405276; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.447972
 CNTKCommandTrainEnd: SVDTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/run-test b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/run-test
index a93f288e5..69ec7918a 100755
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/run-test
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/run-test
@@ -5,11 +5,5 @@
 OriginalTestDir=../../SVD
 ConfigDir=$TEST_DIR/$OriginalTestDir
 
-(cd $TEST_DIR/$OriginalTestDir && md5sum baseline*) | (cd $TEST_DIR && md5sum --status -c -)
-if [ $? != 0 ]; then
-  echo Error: Baselines must match original test.
-  exit 1
-fi
-
 # cntkrun <CNTK config file name> <additional CNTK args>
 cntkrun cntk.cntk 'reader=[readerType=ExperimentalHTKMLFReader] reader=[prefetch=true]' || exit $?
diff --git a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/testcases.yml b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/testcases.yml
index 2ce992112..6ff18c8ef 100644
--- a/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/testcases.yml
+++ b/Tests/EndToEndTests/Speech/ExperimentalHtkmlfReader/SVD/testcases.yml
@@ -8,28 +8,28 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.1%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErr[0]PerSample = {{float,tolerance=.1%}}
 
   SVD decomposition output messages should match:
     patterns:
-      - ^Performing SVD for a   {{integer}}-by-{{integer}}
+      - Performing SVD for a   {{integer}}-by-{{integer}}
diff --git a/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.cpu.txt b/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.cpu.txt
index 823d4f2ae..731afee10 100644
--- a/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.cpu.txt
@@ -8698,4 +8698,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[   5-   5 of 128]: SamplesSeen = 596; TrainLossPerSample =  4.22185009; EvalErr[0]PerSample = 0.91778523; TotalTime = 6.30986s; TotalTimePerSample = 10.58701ms; SamplesPerSecond = 94
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 4.159039; EvalErrPerSample = 0.89682537; AvgLearningRatePerSample = 0.02500000037; EpochTime=32.210996
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.gpu.txt b/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.gpu.txt
index e1eda5cfa..dd4f11701 100644
--- a/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.gpu.txt
@@ -8699,4 +8699,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[   5-   5 of 128]: SamplesSeen = 596; TrainLossPerSample =  4.22185009; EvalErr[0]PerSample = 0.91778523; TotalTime = 4.28988s; TotalTimePerSample = 7.19778ms; SamplesPerSecond = 138
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 4.1590457; EvalErrPerSample = 0.89682537; AvgLearningRatePerSample = 0.02500000037; EpochTime=22.958186
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.windows.cpu.txt
index d8208d2e1..f7b959387 100644
--- a/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.windows.cpu.txt
@@ -8706,4 +8706,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[   5-   5 of 128]: SamplesSeen = 486; TrainLossPerSample =  4.26283798; EvalErr[0]PerSample = 0.87448560; TotalTime = 19.23765s; TotalTimePerSample = 39.58364ms; SamplesPerSecond = 25
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 4.2592373; EvalErrPerSample = 0.87673128; AvgLearningRatePerSample = 0.02500000037; EpochTime=154.44309
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.windows.gpu.txt
index cfc696f43..178df4ec7 100644
--- a/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/LSTM/FullUtterance/baseline.windows.gpu.txt
@@ -8706,4 +8706,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[   5-   5 of 128]: SamplesSeen = 486; TrainLossPerSample =  4.26283999; EvalErr[0]PerSample = 0.87448560; TotalTime = 21.40957s; TotalTimePerSample = 44.05262ms; SamplesPerSecond = 22
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 4.2592301; EvalErrPerSample = 0.87673128; AvgLearningRatePerSample = 0.02500000037; EpochTime=181.15261
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/LSTM/FullUtterance/testcases.yml b/Tests/EndToEndTests/Speech/LSTM/FullUtterance/testcases.yml
index ace281b41..6ae858e17 100644
--- a/Tests/EndToEndTests/Speech/LSTM/FullUtterance/testcases.yml
+++ b/Tests/EndToEndTests/Speech/LSTM/FullUtterance/testcases.yml
@@ -8,24 +8,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.1%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErr[0]PerSample = {{float,tolerance=.1%}}
diff --git a/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/baseline.cpu.txt b/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/baseline.cpu.txt
index 5f0e62efe..50a0ac824 100644
--- a/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/baseline.cpu.txt
@@ -2965,4 +2965,4 @@ CNTKCommandTrainEnd: speechTrain
 
 Action "train" complete.
 
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/baseline.gpu.txt b/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/baseline.gpu.txt
index 95dafc219..25d790bae 100644
--- a/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/baseline.gpu.txt
@@ -2965,4 +2965,4 @@ CNTKCommandTrainEnd: speechTrain
 
 Action "train" complete.
 
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/testcases.yml b/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/testcases.yml
index 066f8d88a..aa7fdd30f 100644
--- a/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/testcases.yml
+++ b/Tests/EndToEndTests/Speech/LSTM/Truncated-Kaldi/testcases.yml
@@ -8,24 +8,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float,tolerance=.01%}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.1%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErr[0]PerSample = {{float,tolerance=.1%}}
diff --git a/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.cpu.txt b/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.cpu.txt
index 4a0909832..8924465d1 100644
--- a/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.cpu.txt
@@ -9212,4 +9212,4 @@ Starting minibatch loop.
  Epoch[ 4 of 4]-Minibatch[  31-  40 of 1024]: SamplesSeen = 1630; TrainLossPerSample =  4.05378643; EvalErr[0]PerSample = 0.89018405; TotalTime = 37.74493s; TotalTimePerSample = 23.15640ms; SamplesPerSecond = 43
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 4.0972614; EvalErrPerSample = 0.86727881; AvgLearningRatePerSample = 0.0007812500116; EpochTime=156.30584
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.gpu.txt b/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.gpu.txt
index 6e9f63c84..cc0d85a0e 100644
--- a/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.gpu.txt
@@ -9213,4 +9213,4 @@ Starting minibatch loop.
  Epoch[ 4 of 4]-Minibatch[  31-  40 of 1024]: SamplesSeen = 1630; TrainLossPerSample =  4.05378643; EvalErr[0]PerSample = 0.89018405; TotalTime = 3.63035s; TotalTimePerSample = 2.22721ms; SamplesPerSecond = 448
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 4.097261; EvalErrPerSample = 0.86727881; AvgLearningRatePerSample = 0.0007812500116; EpochTime=16.609371
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.windows.cpu.txt
index 731e45434..6b8a258f1 100644
--- a/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.windows.cpu.txt
@@ -9221,4 +9221,4 @@ Starting minibatch loop.
  Epoch[ 4 of 4]-Minibatch[  31-  40 of 1024]: SamplesSeen = 1856; TrainLossPerSample =  4.20807937; EvalErr[0]PerSample = 0.90948276; TotalTime = 57.52875s; TotalTimePerSample = 30.99609ms; SamplesPerSecond = 32
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 4.1010361; EvalErrPerSample = 0.86705089; AvgLearningRatePerSample = 0.0007812500116; EpochTime=225.56328
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.windows.gpu.txt
index 102ea6e0c..71cb2f74f 100644
--- a/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/LSTM/Truncated/baseline.windows.gpu.txt
@@ -9221,4 +9221,4 @@ Starting minibatch loop.
  Epoch[ 4 of 4]-Minibatch[  31-  40 of 1024]: SamplesSeen = 1856; TrainLossPerSample =  4.20807937; EvalErr[0]PerSample = 0.90948276; TotalTime = 9.90009s; TotalTimePerSample = 5.33410ms; SamplesPerSecond = 187
 Finished Epoch[ 4 of 4]: [Training Set] TrainLossPerSample = 4.1010361; EvalErrPerSample = 0.86705089; AvgLearningRatePerSample = 0.0007812500116; EpochTime=41.549931
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/LSTM/Truncated/testcases.yml b/Tests/EndToEndTests/Speech/LSTM/Truncated/testcases.yml
index aac59c35f..e56c116ed 100644
--- a/Tests/EndToEndTests/Speech/LSTM/Truncated/testcases.yml
+++ b/Tests/EndToEndTests/Speech/LSTM/Truncated/testcases.yml
@@ -8,24 +8,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float,tolerance=.01%}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.1%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErr[0]PerSample = {{float,tolerance=.1%}}
diff --git a/Tests/EndToEndTests/Speech/QuickE2E/baseline.cpu.txt b/Tests/EndToEndTests/Speech/QuickE2E/baseline.cpu.txt
index 74268f0c3..1ffd26847 100644
--- a/Tests/EndToEndTests/Speech/QuickE2E/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/QuickE2E/baseline.cpu.txt
@@ -340,7 +340,7 @@ Starting minibatch loop.
  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91941869; EvalErr[0]PerSample = 0.52890623; TotalTime = 1.86784s; TotalTimePerSample = 0.18241ms; SamplesPerSecond = 5482
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91062427; EvalErr[0]PerSample = 0.52783203; TotalTime = 1.84987s; TotalTimePerSample = 0.18065ms; SamplesPerSecond = 5535
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9150215; EvalErrPerSample = 0.52836913; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=3.736133
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 running on localhost at 2015/08/25 20:26:31
@@ -710,4 +710,4 @@ Validating --> EvalErrorPrediction = ErrorPrediction(labels[132, 1024], HLast[13
  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91941869; EvalErr[0]PerSample = 0.52890623; TotalTime = 1.88723s; TotalTimePerSample = 0.18430ms; SamplesPerSecond = 5425
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.91062427; EvalErr[0]PerSample = 0.52783203; TotalTime = 1.84469s; TotalTimePerSample = 0.18015ms; SamplesPerSecond = 5551
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9150215; EvalErrPerSample = 0.52836913; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=5.315324
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/QuickE2E/baseline.gpu.txt b/Tests/EndToEndTests/Speech/QuickE2E/baseline.gpu.txt
index d71d3eac2..b2076d6f4 100644
--- a/Tests/EndToEndTests/Speech/QuickE2E/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/QuickE2E/baseline.gpu.txt
@@ -553,7 +553,7 @@ Starting minibatch loop.
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.86656265; EvalErr[0]PerSample = 0.51748047; TotalTime = 0.21814s; TotalTimePerSample = 0.02130ms; SamplesPerSecond = 46943
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.51840824; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=0.493964
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 === Running /home/mluser/src/cplx_master/build/debug/bin/cntk configFile=/home/mluser/src/cplx_master/Tests/Speech/QuickE2E/cntk.config RunDir=/tmp/cntk-test-20151024124900.548963/Speech_QuickE2E@debug_gpu DataDir=/home/mluser/src/cplx_master/Tests/Speech/Data ConfigDir=/home/mluser/src/cplx_master/Tests/Speech/QuickE2E DeviceId=0
@@ -1331,4 +1331,4 @@ EnforceOneGPUOnly: WARNING: Ignored attempt to change GPU choice from 0 now 1. T
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.86656265; EvalErr[0]PerSample = 0.51748047; TotalTime = 0.21717s; TotalTimePerSample = 0.02121ms; SamplesPerSecond = 47152
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.51840824; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=1.439589
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/QuickE2E/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/QuickE2E/baseline.windows.cpu.txt
index ff037b498..f899e6ae1 100644
--- a/Tests/EndToEndTests/Speech/QuickE2E/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/QuickE2E/baseline.windows.cpu.txt
@@ -349,7 +349,7 @@ Starting minibatch loop, distributed reading is: DISABLED
  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.88589633; EvalErr[0]PerSample = 0.52529299; TotalTime = 0.59385s; TotalTimePerSample = 0.05799ms; SamplesPerSecond = 17243
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.89380074; EvalErr[0]PerSample = 0.51816404; TotalTime = 0.51284s; TotalTimePerSample = 0.05008ms; SamplesPerSecond = 19967
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8898486; EvalErrPerSample = 0.52172852; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.109687
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 -------------------------------------------------------------------
@@ -728,4 +728,4 @@ Validating --> EvalErrorPrediction = ErrorPrediction(labels[132, 1024], HLast[13
  Epoch[ 3 of 3]-Minibatch[   1-  10 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.88589633; EvalErr[0]PerSample = 0.52529299; TotalTime = 0.86012s; TotalTimePerSample = 0.08400ms; SamplesPerSecond = 11905
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.89380074; EvalErr[0]PerSample = 0.51816404; TotalTime = 0.60616s; TotalTimePerSample = 0.05919ms; SamplesPerSecond = 16893
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8898486; EvalErrPerSample = 0.52172852; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=2.711551
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/QuickE2E/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/QuickE2E/baseline.windows.gpu.txt
index 0e9622b78..759d87b54 100644
--- a/Tests/EndToEndTests/Speech/QuickE2E/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/QuickE2E/baseline.windows.gpu.txt
@@ -561,7 +561,7 @@ Starting minibatch loop.
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.87358818; EvalErr[0]PerSample = 0.51542969; TotalTime = 0.60551s; TotalTimePerSample = 0.05913ms; SamplesPerSecond = 16911
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8705584; EvalErrPerSample = 0.5186035; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.428405
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 === Running /cygdrive/e/NetScale/CNTK/git_repos/cplx_master2/x64/debug/cntk.exe configFile=E:\NetScale\CNTK\git_repos\cplx_master2\Tests\Speech\QuickE2E/cntk.config RunDir=C:\cygwin64\tmp\cntk-test-20151024140721.927553\Speech_QuickE2E@debug_gpu DataDir=E:\NetScale\CNTK\git_repos\cplx_master2\Tests\Speech\Data ConfigDir=E:\NetScale\CNTK\git_repos\cplx_master2\Tests\Speech\QuickE2E DeviceId=0
@@ -1347,4 +1347,4 @@ Starting minibatch loop.
  Epoch[ 3 of 3]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.87358818; EvalErr[0]PerSample = 0.51542969; TotalTime = 0.86938s; TotalTimePerSample = 0.08490ms; SamplesPerSecond = 11778
 Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8705584; EvalErrPerSample = 0.5186035; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=6.283729
 CNTKCommandTrainEnd: speechTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/QuickE2E/testcases.yml b/Tests/EndToEndTests/Speech/QuickE2E/testcases.yml
index 7d2602bb6..3ffaaa6be 100644
--- a/Tests/EndToEndTests/Speech/QuickE2E/testcases.yml
+++ b/Tests/EndToEndTests/Speech/QuickE2E/testcases.yml
@@ -8,24 +8,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.1%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErr[0]PerSample = {{float,tolerance=.1%}}
diff --git a/Tests/EndToEndTests/Speech/SVD/baseline.cpu.txt b/Tests/EndToEndTests/Speech/SVD/baseline.cpu.txt
index 5c035c099..0dc1812cc 100644
--- a/Tests/EndToEndTests/Speech/SVD/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/SVD/baseline.cpu.txt
@@ -2064,4 +2064,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.85902176; EvalErr[0]PerSample = 0.51396484; TotalTime = 1.48188s; TotalTimePerSample = 0.14471ms; SamplesPerSecond = 6910
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 1.8499603; EvalErrPerSample = 0.5126465; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=3.049746
 CNTKCommandTrainEnd: SVDTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/SVD/baseline.gpu.txt b/Tests/EndToEndTests/Speech/SVD/baseline.gpu.txt
index 779884461..e5315f319 100644
--- a/Tests/EndToEndTests/Speech/SVD/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/SVD/baseline.gpu.txt
@@ -2067,4 +2067,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.81663570; EvalErr[0]PerSample = 0.50869141; TotalTime = 0.23803s; TotalTimePerSample = 0.02325ms; SamplesPerSecond = 43019
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 1.8090889; EvalErrPerSample = 0.504834; AvgLearningRatePerSample = 9.765625146e-05; EpochTime=0.528531
 CNTKCommandTrainEnd: SVDTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/SVD/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/SVD/baseline.windows.cpu.txt
index 3dc50ea83..d6751a433 100644
--- a/Tests/EndToEndTests/Speech/SVD/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/SVD/baseline.windows.cpu.txt
@@ -2073,4 +2073,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.80246696; EvalErr[0]PerSample = 0.50654297; TotalTime = 3.29585s; TotalTimePerSample = 0.32186ms; SamplesPerSecond = 3106
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 1.8185977; EvalErrPerSample = 0.50717777; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=7.151551
 CNTKCommandTrainEnd: SVDTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/SVD/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/SVD/baseline.windows.gpu.txt
index 31c006e0c..7a0b10c09 100644
--- a/Tests/EndToEndTests/Speech/SVD/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/SVD/baseline.windows.gpu.txt
@@ -2075,4 +2075,4 @@ Starting minibatch loop.
  Epoch[ 2 of 2]-Minibatch[  11-  20 of 20]: SamplesSeen = 10240; TrainLossPerSample =  1.78207207; EvalErr[0]PerSample = 0.50322266; TotalTime = 0.63083s; TotalTimePerSample = 0.06160ms; SamplesPerSecond = 16232
 Finished Epoch[ 2 of 2]: [Training Set] TrainLossPerSample = 1.7974522; EvalErrPerSample = 0.50405276; AvgLearningRatePerSample = 9.765625146e-005; EpochTime=1.447972
 CNTKCommandTrainEnd: SVDTrain
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/SVD/testcases.yml b/Tests/EndToEndTests/Speech/SVD/testcases.yml
index 26ce6957c..bc462cdb7 100644
--- a/Tests/EndToEndTests/Speech/SVD/testcases.yml
+++ b/Tests/EndToEndTests/Speech/SVD/testcases.yml
@@ -8,28 +8,28 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.1%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErr[0]PerSample = {{float,tolerance=.1%}}
 
   SVD decomposition output messages should match:
     patterns:
-      - ^Performing SVD for a   {{integer}}-by-{{integer}}
+      - Performing SVD for a   {{integer}}-by-{{integer}}
diff --git a/Tests/EndToEndTests/Speech/Simple/baseline.cpu.txt b/Tests/EndToEndTests/Speech/Simple/baseline.cpu.txt
index 995d6f9b0..9f28cd11d 100644
--- a/Tests/EndToEndTests/Speech/Simple/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Speech/Simple/baseline.cpu.txt
@@ -3602,7 +3602,7 @@ starting epoch 0 at record count 0, and file position 0
 already there from last epoch
 Minibatch[1]: ActualMBSize = 603
 Total Samples Evaluated = 603
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 === Running /home/philly/jenkins/workspace/CNTK-Test-Linux-W1/build/release/bin/cntk configFile=/home/philly/jenkins/workspace/CNTK-Test-Linux-W1/Tests/Speech/Simple/cntk.config RunDir=/tmp/cntk-test-20151119184420.414180/Speech_Simple@release_cpu DataDir=/home/philly/jenkins/workspace/CNTK-Test-Linux-W1/Tests/Speech/Data ConfigDir=/home/philly/jenkins/workspace/CNTK-Test-Linux-W1/Tests/Speech/Simple DeviceId=-1 makeMode=true
@@ -4808,4 +4808,4 @@ starting epoch 0 at record count 0, and file position 0
 already there from last epoch
 Minibatch[1]: ActualMBSize = 603
 Total Samples Evaluated = 603
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/Simple/baseline.gpu.txt b/Tests/EndToEndTests/Speech/Simple/baseline.gpu.txt
index e29dbb303..1c16c780f 100644
--- a/Tests/EndToEndTests/Speech/Simple/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Speech/Simple/baseline.gpu.txt
@@ -3699,7 +3699,7 @@ starting epoch 0 at record count 0, and file position 0
 already there from last epoch
 Minibatch[1]: ActualMBSize = 603
 Total Samples Evaluated = 603
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 === Running /home/philly/jenkins/workspace/CNTK-Test-Linux-W1/build/debug/bin/cntk configFile=/home/philly/jenkins/workspace/CNTK-Test-Linux-W1/Tests/Speech/Simple/cntk.config RunDir=/tmp/cntk-test-20151119193939.119454/Speech_Simple@debug_gpu DataDir=/home/philly/jenkins/workspace/CNTK-Test-Linux-W1/Tests/Speech/Data ConfigDir=/home/philly/jenkins/workspace/CNTK-Test-Linux-W1/Tests/Speech/Simple DeviceId=0 makeMode=true
@@ -4952,4 +4952,4 @@ starting epoch 0 at record count 0, and file position 0
 already there from last epoch
 Minibatch[1]: ActualMBSize = 603
 Total Samples Evaluated = 603
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/Simple/baseline.windows.cpu.txt b/Tests/EndToEndTests/Speech/Simple/baseline.windows.cpu.txt
index 6c931cea2..92fcc4b2b 100644
--- a/Tests/EndToEndTests/Speech/Simple/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Speech/Simple/baseline.windows.cpu.txt
@@ -3613,7 +3613,7 @@ starting epoch 0 at record count 0, and file position 0
 already there from last epoch
 Minibatch[1]: ActualMBSize = 603
 Total Samples Evaluated = 603
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 === Running /cygdrive/c/jenkins/workspace/CNTK-Test-Windows-W1/x64/release/cntk.exe configFile=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\Speech\Simple/cntk.config RunDir=C:\windows\TEMP\cntk-test-20151119141632.280376\Speech_Simple@release_cpu DataDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\Speech\Data ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\Speech\Simple DeviceId=-1 makeMode=true
@@ -4830,4 +4830,4 @@ starting epoch 0 at record count 0, and file position 0
 already there from last epoch
 Minibatch[1]: ActualMBSize = 603
 Total Samples Evaluated = 603
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/Simple/baseline.windows.gpu.txt b/Tests/EndToEndTests/Speech/Simple/baseline.windows.gpu.txt
index 2134a6b93..ef09aa2d2 100644
--- a/Tests/EndToEndTests/Speech/Simple/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Speech/Simple/baseline.windows.gpu.txt
@@ -3710,7 +3710,7 @@ starting epoch 0 at record count 0, and file position 0
 already there from last epoch
 Minibatch[1]: ActualMBSize = 603
 Total Samples Evaluated = 603
-COMPLETED
+__COMPLETED__
 === Deleting last epoch data
 ==== Re-running from checkpoint
 === Running /cygdrive/c/jenkins/workspace/CNTK-Test-Windows-W1/x64/debug/cntk.exe configFile=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\Speech\Simple/cntk.config RunDir=C:\Windows\TEMP\cntk-test-20151119120636.543625\Speech_Simple@debug_gpu DataDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\Speech\Data ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\Speech\Simple DeviceId=0 makeMode=true
@@ -4974,4 +4974,4 @@ starting epoch 0 at record count 0, and file position 0
 already there from last epoch
 Minibatch[1]: ActualMBSize = 603
 Total Samples Evaluated = 603
-COMPLETED
+__COMPLETED__
diff --git a/Tests/EndToEndTests/Speech/Simple/testcases.yml b/Tests/EndToEndTests/Speech/Simple/testcases.yml
index 7d2602bb6..3ffaaa6be 100644
--- a/Tests/EndToEndTests/Speech/Simple/testcases.yml
+++ b/Tests/EndToEndTests/Speech/Simple/testcases.yml
@@ -8,24 +8,24 @@ tags:
 testCases:
   CNTK Run must be completed:
     patterns:
-      - ^COMPLETED
+      - __COMPLETED__
 
   Must train epochs in exactly same order and parameters:
     patterns:
-      - ^Starting Epoch {{integer}}
+      - Starting Epoch {{integer}}
       - learning rate per sample = {{float}}
       - momentum = {{float}}
 
   Epochs must be finished with expected results:
     patterns:
-      - ^Finished Epoch[{{integer}} of {{integer}}]
+      - Finished Epoch[{{integer}} of {{integer}}]
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErrPerSample = {{float,tolerance=.1%}}
       - AvgLearningRatePerSample = {{float,tolerance=0.001%}}
 
   Per-minibatch training results must match:
     patterns:
-      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
+      - Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}}
       - SamplesSeen = {{integer}}
       - TrainLossPerSample = {{float,tolerance=.1%}}
       - EvalErr[0]PerSample = {{float,tolerance=.1%}}
diff --git a/Tests/EndToEndTests/Text/SparseDSSM/baseline.cpu.txt b/Tests/EndToEndTests/Text/SparseDSSM/baseline.cpu.txt
index ba8e90eb8..0498a9f5e 100755
--- a/Tests/EndToEndTests/Text/SparseDSSM/baseline.cpu.txt
+++ b/Tests/EndToEndTests/Text/SparseDSSM/baseline.cpu.txt
@@ -471,7 +471,7 @@ MPI Rank 0: Allocating matrices for forward and/or backward propagation.
 MPI Rank 0: Final Results: Minibatch[1-25]: Samples Seen = 102399    CE: CrossEntropyWithSoftmax/Sample = 1.6991001    Perplexity = 5.4690235    
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.6991001
 MPI Rank 0: CNTKCommandTrainEnd: train
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -850,7 +850,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.74922504; TotalTime = 7.2778s; SamplesPerSecond = 1407.0
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.7563989; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=18.6685
 MPI Rank 1: CNTKCommandTrainEnd: train
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -1229,7 +1229,7 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.76783981; TotalTime = 7.2778s; SamplesPerSecond = 1407.0
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.7563989; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=18.6685
 MPI Rank 2: CNTKCommandTrainEnd: train
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: -------------------------------------------------------------------
 MPI Rank 3: Build info: 
@@ -1608,7 +1608,7 @@ MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.73868065; TotalTime = 7.2778s; SamplesPerSecond = 1407.0
 MPI Rank 3: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.7563989; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=18.6685
 MPI Rank 3: CNTKCommandTrainEnd: train
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
 === Deleting last epoch data
 ==== Re-running from checkpoint
@@ -2065,7 +2065,7 @@ MPI Rank 0: Allocating matrices for forward and/or backward propagation.
 MPI Rank 0: Final Results: Minibatch[1-25]: Samples Seen = 102399    CE: CrossEntropyWithSoftmax/Sample = 1.7629802    Perplexity = 5.8297857    
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.7629802
 MPI Rank 0: CNTKCommandTrainEnd: train
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -2434,7 +2434,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.80574760; TotalTime = 6.4066s; SamplesPerSecond = 1598.4
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8025137; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=18.3486
 MPI Rank 1: CNTKCommandTrainEnd: train
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -2803,7 +2803,7 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.82326355; TotalTime = 6.4066s; SamplesPerSecond = 1598.4
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8025137; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=18.3486
 MPI Rank 2: CNTKCommandTrainEnd: train
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: -------------------------------------------------------------------
 MPI Rank 3: Build info: 
@@ -3172,5 +3172,5 @@ MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.79679737; TotalTime = 6.4066s; SamplesPerSecond = 1598.4
 MPI Rank 3: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8025137; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=18.3486
 MPI Rank 3: CNTKCommandTrainEnd: train
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Text/SparseDSSM/baseline.gpu.txt b/Tests/EndToEndTests/Text/SparseDSSM/baseline.gpu.txt
index b8d382b91..60cb26a35 100755
--- a/Tests/EndToEndTests/Text/SparseDSSM/baseline.gpu.txt
+++ b/Tests/EndToEndTests/Text/SparseDSSM/baseline.gpu.txt
@@ -472,7 +472,7 @@ MPI Rank 0: Allocating matrices for forward and/or backward propagation.
 MPI Rank 0: Final Results: Minibatch[1-25]: Samples Seen = 102399    CE: CrossEntropyWithSoftmax/Sample = 1.8091086    Perplexity = 6.1050027    
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.8091086
 MPI Rank 0: CNTKCommandTrainEnd: train
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -852,7 +852,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.86598778; TotalTime = 4.5356s; SamplesPerSecond = 2257.7
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8856394; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=11.692
 MPI Rank 1: CNTKCommandTrainEnd: train
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -1232,7 +1232,7 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.87902641; TotalTime = 4.5357s; SamplesPerSecond = 2257.7
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8856394; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=11.692
 MPI Rank 2: CNTKCommandTrainEnd: train
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: -------------------------------------------------------------------
 MPI Rank 3: Build info: 
@@ -1612,7 +1612,7 @@ MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.88304176; TotalTime = 4.5357s; SamplesPerSecond = 2257.6
 MPI Rank 3: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8856394; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=11.692
 MPI Rank 3: CNTKCommandTrainEnd: train
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
 === Deleting last epoch data
 ==== Re-running from checkpoint
@@ -2069,7 +2069,7 @@ MPI Rank 0: Allocating matrices for forward and/or backward propagation.
 MPI Rank 0: Final Results: Minibatch[1-25]: Samples Seen = 102399    CE: CrossEntropyWithSoftmax/Sample = 1.8982234    Perplexity = 6.6740266    
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.8982234
 MPI Rank 0: CNTKCommandTrainEnd: train
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: -------------------------------------------------------------------
 MPI Rank 1: Build info: 
@@ -2438,7 +2438,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.94722424; TotalTime = 4.5333s; SamplesPerSecond = 2258.9
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9504802; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=12.2322
 MPI Rank 1: CNTKCommandTrainEnd: train
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: -------------------------------------------------------------------
 MPI Rank 2: Build info: 
@@ -2807,7 +2807,7 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.95635300; TotalTime = 4.5332s; SamplesPerSecond = 2258.9
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9504802; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=12.2322
 MPI Rank 2: CNTKCommandTrainEnd: train
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: -------------------------------------------------------------------
 MPI Rank 3: Build info: 
@@ -3176,5 +3176,5 @@ MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.95965519; TotalTime = 4.5333s; SamplesPerSecond = 2258.9
 MPI Rank 3: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9504802; AvgLearningRatePerSample = 9.9999997e-05; EpochTime=12.2322
 MPI Rank 3: CNTKCommandTrainEnd: train
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Text/SparseDSSM/baseline.windows.cpu.txt b/Tests/EndToEndTests/Text/SparseDSSM/baseline.windows.cpu.txt
index c0d0efb12..34aa4ff23 100755
--- a/Tests/EndToEndTests/Text/SparseDSSM/baseline.windows.cpu.txt
+++ b/Tests/EndToEndTests/Text/SparseDSSM/baseline.windows.cpu.txt
@@ -483,7 +483,7 @@ MPI Rank 0: Allocating matrices for forward and/or backward propagation.
 MPI Rank 0: Final Results: Minibatch[1-25]: Samples Seen = 102399    CE: CrossEntropyWithSoftmax/Sample = 1.8119593    Perplexity = 6.1224311    
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.8119593
 MPI Rank 0: CNTKCommandTrainEnd: train
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303160843.247203\Text_SparseDSSM@release_cpu/stderr_train.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -866,7 +866,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.86772957; TotalTime = 16.2062s; SamplesPerSecond = 631.9
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8883394; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=40.2826
 MPI Rank 1: CNTKCommandTrainEnd: train
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303160843.247203\Text_SparseDSSM@release_cpu/stderr_train.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1249,7 +1249,7 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.88702583; TotalTime = 15.9308s; SamplesPerSecond = 642.8
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8883394; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=40.2866
 MPI Rank 2: CNTKCommandTrainEnd: train
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303160843.247203\Text_SparseDSSM@release_cpu/stderr_train.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -1632,7 +1632,7 @@ MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.87299423; TotalTime = 16.2062s; SamplesPerSecond = 631.9
 MPI Rank 3: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8883394; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=40.2823
 MPI Rank 3: CNTKCommandTrainEnd: train
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
 === Deleting last epoch data
 ==== Re-running from checkpoint
@@ -2101,7 +2101,7 @@ MPI Rank 0: Allocating matrices for forward and/or backward propagation.
 MPI Rank 0: Final Results: Minibatch[1-25]: Samples Seen = 102399    CE: CrossEntropyWithSoftmax/Sample = 1.8993035    Perplexity = 6.6812396    
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.8993035
 MPI Rank 0: CNTKCommandTrainEnd: train
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303160843.247203\Text_SparseDSSM@release_cpu/stderr_train.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -2474,7 +2474,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.94770432; TotalTime = 17.5215s; SamplesPerSecond = 584.4
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9528804; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=46.4912
 MPI Rank 1: CNTKCommandTrainEnd: train
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303160843.247203\Text_SparseDSSM@release_cpu/stderr_train.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -2847,7 +2847,7 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.96577854; TotalTime = 17.2169s; SamplesPerSecond = 594.8
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9528804; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=46.4944
 MPI Rank 2: CNTKCommandTrainEnd: train
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303160843.247203\Text_SparseDSSM@release_cpu/stderr_train.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -3220,5 +3220,5 @@ MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.95135975; TotalTime = 17.5215s; SamplesPerSecond = 584.4
 MPI Rank 3: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9528804; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=46.4913
 MPI Rank 3: CNTKCommandTrainEnd: train
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/Text/SparseDSSM/baseline.windows.gpu.txt b/Tests/EndToEndTests/Text/SparseDSSM/baseline.windows.gpu.txt
index cc2c98ca8..974a065a1 100755
--- a/Tests/EndToEndTests/Text/SparseDSSM/baseline.windows.gpu.txt
+++ b/Tests/EndToEndTests/Text/SparseDSSM/baseline.windows.gpu.txt
@@ -484,7 +484,7 @@ MPI Rank 0: Allocating matrices for forward and/or backward propagation.
 MPI Rank 0: Final Results: Minibatch[1-25]: Samples Seen = 102399    CE: CrossEntropyWithSoftmax/Sample = 1.8091086    Perplexity = 6.1050028    
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.8091086
 MPI Rank 0: CNTKCommandTrainEnd: train
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303154710.317558\Text_SparseDSSM@release_gpu/stderr_train.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -868,7 +868,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.86598778; TotalTime = 8.0798s; SamplesPerSecond = 1267.4
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8856394; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=20.8703
 MPI Rank 1: CNTKCommandTrainEnd: train
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303154710.317558\Text_SparseDSSM@release_gpu/stderr_train.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -1252,7 +1252,7 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.87902641; TotalTime = 8.0798s; SamplesPerSecond = 1267.4
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8856394; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=20.8703
 MPI Rank 2: CNTKCommandTrainEnd: train
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303154710.317558\Text_SparseDSSM@release_gpu/stderr_train.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -1636,7 +1636,7 @@ MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.88304138; TotalTime = 8.0823s; SamplesPerSecond = 1267.0
 MPI Rank 3: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.8856394; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=20.8703
 MPI Rank 3: CNTKCommandTrainEnd: train
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
 === Deleting last epoch data
 ==== Re-running from checkpoint
@@ -2105,7 +2105,7 @@ MPI Rank 0: Allocating matrices for forward and/or backward propagation.
 MPI Rank 0: Final Results: Minibatch[1-25]: Samples Seen = 102399    CE: CrossEntropyWithSoftmax/Sample = 1.8982234    Perplexity = 6.6740265    
 MPI Rank 0: Finished Epoch[ 3 of 3]: [Validation Set] TrainLossPerSample = 1.8982234
 MPI Rank 0: CNTKCommandTrainEnd: train
-MPI Rank 0: COMPLETED
+MPI Rank 0: __COMPLETED__
 MPI Rank 0: ~MPIWrapper
 MPI Rank 1: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303154710.317558\Text_SparseDSSM@release_gpu/stderr_train.logrank1
 MPI Rank 1: -------------------------------------------------------------------
@@ -2478,7 +2478,7 @@ MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 1:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.94722424; TotalTime = 8.1200s; SamplesPerSecond = 1261.1
 MPI Rank 1: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9504802; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=22.065
 MPI Rank 1: CNTKCommandTrainEnd: train
-MPI Rank 1: COMPLETED
+MPI Rank 1: __COMPLETED__
 MPI Rank 1: ~MPIWrapper
 MPI Rank 2: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303154710.317558\Text_SparseDSSM@release_gpu/stderr_train.logrank2
 MPI Rank 2: -------------------------------------------------------------------
@@ -2851,7 +2851,7 @@ MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 2:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.95635300; TotalTime = 8.1200s; SamplesPerSecond = 1261.1
 MPI Rank 2: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9504802; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=22.0648
 MPI Rank 2: CNTKCommandTrainEnd: train
-MPI Rank 2: COMPLETED
+MPI Rank 2: __COMPLETED__
 MPI Rank 2: ~MPIWrapper
 MPI Rank 3: Redirecting stderr to file C:\cygwin64\tmp\cntk-test-20160303154710.317558\Text_SparseDSSM@release_gpu/stderr_train.logrank3
 MPI Rank 3: -------------------------------------------------------------------
@@ -3224,5 +3224,5 @@ MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[   1-  10, 40.00%]: SamplesSeen = 10240; T
 MPI Rank 3:  Epoch[ 3 of 3]-Minibatch[  11-  20, 80.00%]: SamplesSeen = 10240; TrainLossPerSample =  1.95965519; TotalTime = 8.1200s; SamplesPerSecond = 1261.1
 MPI Rank 3: Finished Epoch[ 3 of 3]: [Training Set] TrainLossPerSample = 1.9504802; AvgLearningRatePerSample = 9.9999997e-005; EpochTime=22.065
 MPI Rank 3: CNTKCommandTrainEnd: train
-MPI Rank 3: COMPLETED
+MPI Rank 3: __COMPLETED__
 MPI Rank 3: ~MPIWrapper
diff --git a/Tests/EndToEndTests/run-test-common b/Tests/EndToEndTests/run-test-common
index 93047823c..c4fe0a5c0 100755
--- a/Tests/EndToEndTests/run-test-common
+++ b/Tests/EndToEndTests/run-test-common
@@ -71,7 +71,7 @@ cntkrun()
     OutputDir=$(cygpath -aw $OutputDir)
   fi
 
-  CNTKArgs="configFile=$ConfigDir/$configFileName currentDirectory=$DataDir RunDir=$RunDir DataDir=$DataDir ConfigDir=$ConfigDir OutputDir=$OutputDir DeviceId=$CNTKDeviceId $additionalCNTKArgs"
+  CNTKArgs="configFile=$ConfigDir/$configFileName currentDirectory=$DataDir RunDir=$RunDir DataDir=$DataDir ConfigDir=$ConfigDir OutputDir=$OutputDir DeviceId=$CNTKDeviceId timestamping=true $additionalCNTKArgs"
   if [ "$LogFileName" != "" ]; then
     CNTKArgs="$CNTKArgs stderr=$RunDir/$LogFileName"
   fi