Integrate 2eaef84d7b into master

Examples/Image/Classification/ConvNet/BrainScript/ConvNetLRN_CIFAR10_DataAug.cntk

@@ -32,7 +32,7 @@ TrainConvNet = {
                 x2s = SplitDimension(x2, 3, 1) 
                 # 3D convolution with a filter that has a non 1-size only in the 3rd axis, and does not reduce since the reduction dimension is fake and 1
                 W = ParameterTensor{(1:1:2*n+1:1), learningRateMultiplier = 0, initValue = alpha/(2*n+1)}
-                y = Convolution (W, x2s, (1:1:2*n+1), mapDims = 1, stride = 1, sharing = true, autoPadding = true, lowerPad = 0, upperPad = 0, transpose = false, maxTempMemSizeInSamples = 0)
+                y = Convolution (W, x2s, (1:1:2*n+1), mapDims = 1, stride = 1, sharing = true, autoPadding = true, lowerPad = 0, upperPad = 0, maxTempMemSizeInSamples = 0)
                 # reshape back to remove the fake singleton reduction dimension
                 b = FlattenDimensions(y, 3, 2)
                 den = Exp (beta .* Log(k + b)) 

Examples/Image/Classification/ConvNet/Python/ConvNetLRN_CIFAR10_DataAug.py

@@ -117,7 +117,7 @@ def convnetlrn_cifar10_dataaug(reader_train, reader_test, epoch_size=50000, max_
     }
 
     cntk.utils.log_number_of_parameters(z) ; print()
-    progress_printer = cntk.utils.ProgressPrinter(tag='Training')
+    progress_printer = cntk.utils.ProgressPrinter(tag='Training', num_epochs=max_epochs)
 
     # perform model training
     for epoch in range(max_epochs):       # loop over epochs

Examples/Image/Classification/ConvNet/Python/ConvNet_CIFAR10.py

@@ -84,10 +84,10 @@ def convnet_cifar10(debug_output=False):
     }
 
     cntk.utils.log_number_of_parameters(z) ; print()
-    progress_printer = cntk.utils.ProgressPrinter(tag='Training')
+    max_epochs = 30
+    progress_printer = cntk.utils.ProgressPrinter(tag='Training', num_epochs=max_epochs)
 
     # Get minibatches of images to train with and perform model training
-    max_epochs = 30
     for epoch in range(max_epochs):       # loop over epochs
         sample_count = 0
         while sample_count < epoch_size:  # loop over minibatches in the epoch

Examples/Image/Classification/ConvNet/Python/ConvNet_CIFAR10_DataAug.py

@@ -95,7 +95,7 @@ def convnet_cifar10_dataaug(reader_train, reader_test, epoch_size = 50000, max_e
     }
 
     cntk.utils.log_number_of_parameters(z) ; print()
-    progress_printer = cntk.utils.ProgressPrinter(tag='Training')
+    progress_printer = cntk.utils.ProgressPrinter(tag='Training', num_epochs=max_epochs)
 
     # perform model training
     for epoch in range(max_epochs):       # loop over epochs

Examples/Image/Classification/ConvNet/Python/ConvNet_MNIST.py

@@ -74,10 +74,10 @@ def convnet_mnist(debug_output=False):
     }
 
     cntk.utils.log_number_of_parameters(z) ; print()
-    progress_printer = cntk.utils.ProgressPrinter(tag='Training')
+    max_epochs = 40
+    progress_printer = cntk.utils.ProgressPrinter(tag='Training', num_epochs=max_epochs)
 
     # Get minibatches of images to train with and perform model training
-    max_epochs = 40
     for epoch in range(max_epochs):       # loop over epochs
         sample_count = 0
         while sample_count < epoch_size:  # loop over minibatches in the epoch

Examples/Image/Classification/ResNet/Python/TrainResNet_CIFAR10.py

@@ -97,7 +97,7 @@ def train_and_evaluate(reader_train, reader_test, network_name, epoch_size, max_
     }
 
     log_number_of_parameters(z) ; print()
-    progress_printer = ProgressPrinter(tag='Training')
+    progress_printer = ProgressPrinter(tag='Training', num_epochs=max_epochs)
 
     # perform model training
     

Examples/Image/DataSets/Grocery/install_grocery.py

@@ -7,6 +7,9 @@
 from __future__ import print_function
 import zipfile
 import os
+from sys import platform
+import shutil
+
 try:
     from urllib.request import urlretrieve 
 except ImportError: 
@@ -26,6 +29,15 @@ def download_grocery_data():
             print('Extracting ' + filename + '...')
             with zipfile.ZipFile(filename) as myzip:
                 myzip.extractall(dataset_folder)
+            if platform != "win32":
+                testfile  = os.path.join(dataset_folder, "grocery", "test.txt")
+                unixfile = os.path.join(dataset_folder, "grocery", "test_unix.txt")
+                out = open(unixfile, 'w')
+                with open(testfile) as f:
+                    for line in f:
+                        out.write(line.replace('\\', '/'))
+                out.close()
+                shutil.move(unixfile, testfile)
         finally:
             os.remove(filename)
         print('Done.')
@@ -34,4 +46,4 @@ def download_grocery_data():
     
 if __name__ == "__main__":
     download_grocery_data()
-    
+    

Examples/LanguageUnderstanding/ATIS/Python/LanguageUnderstanding.py

@@ -13,8 +13,7 @@ import cntk
 # variables and stuff  #
 ########################
 
-cntk_dir = os.path.dirname(os.path.abspath(__file__)) + "/../../../.."  # data resides in the CNTK folder
+data_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "Data")
-data_dir = cntk_dir + "/Examples/LanguageUnderstanding/ATIS/Data"       # under Examples/LanguageUnderstanding/ATIS
 vocab_size = 943 ; num_labels = 129 ; num_intents = 26    # number of words in vocab, slot labels, and intent labels
 
 model_dir = "./Models"
@@ -92,8 +91,8 @@ def train(reader, model, max_epochs):
 
     # process minibatches and perform model training
     cntk.utils.log_number_of_parameters(z) ; print()
-    progress_printer = cntk.ProgressPrinter(freq=100, first=10, tag='Training') # more detailed logging
+    progress_printer = cntk.ProgressPrinter(freq=100, first=10, tag='Training', num_epochs=max_epochs) # more detailed logging
-    #progress_printer = ProgressPrinter(tag='Training')
+    #progress_printer = ProgressPrinter(tag='Training', num_epochs=max_epochs)
 
     t = 0
  

Examples/Text/CharacterLM/char_rnn.py

@@ -139,7 +139,7 @@ def create_inputs(vocab_dim):
     return input_sequence, label_sequence
 
 # Creates and trains a character-level language model
-def train_lm(training_file, max_num_minibatches):
+def train_lm(training_file, epochs, max_num_minibatches):
 
     # load the data and vocab
     data, char_to_ix, ix_to_char, data_size, vocab_dim = load_data_and_vocab(training_file)
@@ -168,46 +168,34 @@ def train_lm(training_file, max_num_minibatches):
     trainer = Trainer(z, (ce, errs), learner)
 
     sample_freq = 1000
-    epochs = 50
+    minibatches_per_epoch = min(data_size // minibatch_size, max_num_minibatches // epochs)
-    minibatches_per_epoch = int((data_size / minibatch_size))
-    minibatches = min(epochs * minibatches_per_epoch, max_num_minibatches)
 
     # print out some useful training information
-    log_number_of_parameters(z) ; print()
+    log_number_of_parameters(z)
-    progress_printer = ProgressPrinter(freq=100, tag='Training')    
+    print ("Running %d epochs with %d minibatches per epoch" % (epochs, minibatches_per_epoch))
+    print()
     
-    e = 0
+    progress_printer = ProgressPrinter(freq=100, tag='Training')
-    p = 0
-    for i in range(0, minibatches):
-
-        if p + minibatch_size+1 >= data_size:
-            p = 0
-            e += 1
-            model_filename = "models/shakespeare_epoch%d.dnn" % e
-            z.save(model_filename)
-            print("Saved model to '%s'" % model_filename)
-
-        # get the data            
-        features, labels = get_data(p, minibatch_size, data, char_to_ix, vocab_dim)
 
+    for e in range(0, epochs):
         # Specify the mapping of input variables in the model to actual minibatch data to be trained with
         # If it's the start of the data, we specify that we are looking at a new sequence (True)
-        mask = [False] 
+        mask = [True]
-        if p == 0:
+        for b in range(0, minibatches_per_epoch):
-            mask = [True]
+            # get the data            
-        arguments = ({input_sequence : features, label_sequence : labels}, mask)
+            features, labels = get_data(b, minibatch_size, data, char_to_ix, vocab_dim)
-        trainer.train_minibatch(arguments)
+            arguments = ({input_sequence : features, label_sequence : labels}, mask)
+            mask = [False] 
+            trainer.train_minibatch(arguments)
 
-        progress_printer.update_with_trainer(trainer, with_metric=True) # log progress
+            progress_printer.update_with_trainer(trainer, with_metric=True) # log progress
-        
+            global_minibatch = e*minibatches_per_epoch + b
-        if i % sample_freq == 0:
+            if global_minibatch % sample_freq == 0:
-            print(sample(z, ix_to_char, vocab_dim, char_to_ix))
+                print(sample(z, ix_to_char, vocab_dim, char_to_ix))
 
-        p += minibatch_size
+        model_filename = "models/shakespeare_epoch%d.dnn" % (e+1)
-
+        z.save_model(model_filename)
-    # Do a final save of the model        
+        print("Saved model to '%s'" % model_filename)
-    model_filename = "models/shakespeare_epoch%d.dnn" % e
-    z.save(model_filename)
 
 
 def load_and_sample(model_filename, vocab_filename, prime_text='', use_hardmax=False, length=1000, temperature=1.0):
@@ -223,13 +211,13 @@ def load_and_sample(model_filename, vocab_filename, prime_text='', use_hardmax=F
         
     return sample(model, ix_to_char, len(chars), char_to_ix, prime_text=prime_text, use_hardmax=use_hardmax, length=length, temperature=temperature)
     
-def train_and_eval_char_rnn(max_num_minibatches=sys.maxsize):
+def train_and_eval_char_rnn(epochs=50, max_num_minibatches=sys.maxsize):
-    # train the LM    
+    # train the LM 
-    train_lm("data/tinyshakespeare.txt", max_num_minibatches)
+    train_lm("data/tinyshakespeare.txt", epochs, max_num_minibatches)
 
     # load and sample
     text = "T"
-    return load_and_sample("models/shakespeare_epoch0.dnn", "data/tinyshakespeare.txt.vocab", prime_text=text, use_hardmax=False, length=100, temperature=0.95)
+    return load_and_sample("models/shakespeare_epoch%d.dnn" % (epochs), "data/tinyshakespeare.txt.vocab", prime_text=text, use_hardmax=False, length=100, temperature=0.95)
 
 if __name__=='__main__':    
     # Specify the target device to be used for computing, if you do not want to

Examples/Video/GettingStarted/Python/Conv3D_UCF11.py

@@ -23,7 +23,7 @@ from _cntk_py import set_computation_network_trace_level
 
 # Paths relative to current python file.
 abs_path   = os.path.dirname(os.path.abspath(__file__))
-data_path  = os.path.join(abs_path, "..", "..", "Datasets", "UCF11")
+data_path  = os.path.join(abs_path, "..", "..", "DataSets", "UCF11")
 model_path = os.path.join(abs_path, "Models")
 
 # Define the reader for both training and evaluation action.
@@ -194,14 +194,14 @@ def conv3d_ucf11(train_reader, test_reader, max_epochs=30):
     lr_per_sample          = [0.01]*10+[0.001]*10+[0.0001]
     lr_schedule            = learning_rate_schedule(lr_per_sample, epoch_size=epoch_size, unit=UnitType.sample)
     momentum_time_constant = 4096
-    mm_schedule            = momentum_as_time_constant_schedule(momentum_time_constant, epoch_size=epoch_size)
+    mm_schedule            = momentum_as_time_constant_schedule([momentum_time_constant], epoch_size=epoch_size)
 
     # Instantiate the trainer object to drive the model training
     learner     = momentum_sgd(z.parameters, lr_schedule, mm_schedule, True)
     trainer     = Trainer(z, (ce, pe), learner)
 
     log_number_of_parameters(z) ; print()
-    progress_printer = ProgressPrinter(tag='Training')
+    progress_printer = ProgressPrinter(tag='Training', num_epochs=max_epochs)
 
     # Get minibatches of images to train with and perform model training
     for epoch in range(max_epochs):       # loop over epochs

Source/Math/GPUMatrix.cu

@@ -2374,7 +2374,9 @@ ElemType GPUMatrix<ElemType>::AbsoluteMax() const
         int resInd = 0;
         cublasIdamax(cuHandle, (CUDA_LONG)GetNumElements(), reinterpret_cast<double*>(Data()), 1, &resInd);
         resInd--;
+
         CUDA_CALL(cudaMemcpy(reinterpret_cast<double*>(&res), Data() + resInd, sizeof(double), cudaMemcpyDeviceToHost));
+
         return res;
     }
 }

Tests/EndToEndTests/CNTKv2Python/Examples/char_rnn_test.py

@@ -17,5 +17,5 @@ def test_char_rnn(device_id):
     set_default_device(cntk_device(device_id))
 
     # Just run and verify it does not crash
-    output = train_and_eval_char_rnn(200)
+    output = train_and_eval_char_rnn(1, 200)
     print(output)

Tutorials/CNTK_201B_CIFAR-10_ImageHandsOn.ipynb

@@ -483,7 +483,7 @@
     "    }\n",
     "\n",
     "    log_number_of_parameters(z) ; print()\n",
-    "    progress_printer = ProgressPrinter(tag='Training')\n",
+    "    progress_printer = ProgressPrinter(tag='Training', num_epochs=max_epochs)\n",
     "\n",
     "    # perform model training\n",
     "    batch_index = 0\n",

Tutorials/CNTK_202_Language_Understanding.ipynb

@@ -437,8 +437,8 @@
     "\n",
     "    # process minibatches and perform model training\n",
     "    log_number_of_parameters(model)\n",
-    "    progress_printer = ProgressPrinter(tag='Training')\n",
+    "    progress_printer = ProgressPrinter(tag='Training', num_epochs=max_epochs)\n",
-    "    #progress_printer = ProgressPrinter(freq=100, first=10, tag='Training') # more detailed logging\n",
+    "    #progress_printer = ProgressPrinter(freq=100, first=10, tag='Training', num_epochs=max_epochs) # more detailed logging\n",
     "\n",
     "    t = 0\n",
     "    for epoch in range(max_epochs):         # loop over epochs\n",
@@ -560,7 +560,7 @@
     "    dummy_learner = adam_sgd(criterion.parameters, \n",
     "                             lr=lr_schedule, momentum=momentum_as_time_constant, low_memory=True)\n",
     "    evaluator = Trainer(model, criterion, dummy_learner)\n",
-    "    progress_printer = ProgressPrinter(tag='Evaluation')\n",
+    "    progress_printer = ProgressPrinter(tag='Evaluation', num_epochs=0)\n",
     "\n",
     "    while True:\n",
     "        minibatch_size = 500\n",