9fc1c45e16
Adds native exception catching and rethrowing to managed layer.
...
The managed wrapper provides custom managed exceptions mapping to the native exceptions thrown (runtime_error, logic_error, invalid_argument, and bad_alloc). Other exceptions are thrown in the base custom exception class CNTKException.
Adds code for reporting, during an exception, the stack trace (as a string and as part of the exception message) to the managed layer from the native layer.
MathTests now uses the shared DebugUtil.cpp file instead of its own copy.
2016-02-17 15:15:40 +01:00
bc44e2d212
TensorOp() now verifies that inputs are not sparse;
...
bug fixes in BatchSequenceReader::GetMinibatch(): must use correct dimension in case of output labelType = 'none', now also supports DENSE matrices
2016-02-14 22:57:06 -08:00
03e8cc5266
Addressed code review feedback.
2016-02-12 09:34:17 -08:00
958f634fd4
Fixed builds.
2016-02-12 09:33:18 -08:00
11fe2c64b0
Added eps and engine parameters to BN node.
2016-02-12 09:32:49 -08:00
9ad6147e80
Batch norm: added epsilon as parameter, enabled run mean/invstddev.
2016-02-12 09:32:35 -08:00
1379f58564
Added convolutional implementation for batch norm.
2016-02-12 09:32:21 -08:00
5846fae253
Improved BN unit tests.
2016-02-12 09:32:07 -08:00
21575f3618
Added batch norm per-activation backprop implementation.
2016-02-12 09:31:53 -08:00
4e9483b0ac
Added batch norm per-activation feed forward implementation.
2016-02-12 09:31:39 -08:00
6465c8018c
Integrate mahilleb/02_Convolution_Unstable into master
2016-02-12 08:14:42 -08:00
4e8c3e7edc
Tests/EndToEndTests/Examples/Image/MNIST/02_Convolution: unstable on Windows
2016-02-12 16:44:58 +01:00
7d248101a3
Integrate mahilleb/TestDriverWinCpuonly into master
2016-02-12 07:01:12 -08:00
3021ad016a
Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations
2016-02-10 11:53:23 -08:00
4a3b6ea266
Tests/TestDriver.py: pick up Windows CPU-only binaries from x64/{Debug,Release}_CpuOnly
2016-02-10 17:52:49 +01:00
1b29d2fd69
.vcxproj: let intermediate output go into a common directory
...
Fixes #12 .
2016-02-10 09:10:07 +01:00
2c2a721fa8
CNTK.Cpp.props: define properties to distinguish build flavors and use them
2016-02-10 09:05:31 +01:00
ea55623979
Introduce CNTK.Cpp.props and load in every .vcxproj
2016-02-10 09:05:31 +01:00
52aad32084
.vcxproj: drop include path $(VCInstallDir)atlmfc\include
2016-02-10 09:05:31 +01:00
3b979f877a
.vcxproj: coalesce importing for .user.props for Debug / Release
2016-02-10 09:05:31 +01:00
5a67cf23b7
Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations
2016-02-08 13:10:49 -08:00
c214980850
Disabled HTKMLFReader unit tests HTKMLFReaderSimpleDataLoop7 and HTKMLFReaderSimpleDataLoop17 as they use invalid config setting combination of Truncated=true and frameMode=true
2016-02-08 13:10:40 -08:00
bb608f813e
LSTM test now uses an actual 1D vector for the bias parameter (as opposed to a 1-column 2D matrix)
2016-02-06 20:31:48 -08:00
f3216254fd
Added a check for inconsistent frameMode and Truncated settings. Also addresed some minor CR feedback.
2016-02-05 23:06:15 -08:00
7d1fc361a9
fixed .config to .cntk in CSEvalCient
2016-02-04 09:11:06 +01:00
7ab3d9e909
.vcxproj: Let diagnostics about dependencies point to CNTK Wiki
2016-02-03 17:22:13 +01:00
d23eecc7b1
Fixed TestDriver.py output format regressions introduced in dc1ed4628f
2016-02-03 09:28:17 +01:00
242d563e4a
Integrate amitaga/devicePlacement into master
2016-02-02 15:57:41 -08:00
d8b0c70e02
Merge pull request #60 from stonebig/patch-7
...
python 3 compatibility fix
2016-02-02 14:10:58 -08:00
0dbd36e88f
Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/devicePlacement
2016-02-02 13:14:59 -08:00
b0dcec6c14
removing models files after e2e test runs
2016-02-02 20:08:31 +01:00
ee87f741b0
Added VS debugging command line args for example e2es
2016-02-02 20:08:31 +01:00
7e424021f5
moving example e2es to E leg, adding all to VS solution
2016-02-02 20:08:31 +01:00
244d351bbf
Fixing wrong name in the linking
2016-02-02 14:42:08 +01:00
b9d8961378
Fixing compiler issues + renaming
2016-02-02 14:34:21 +01:00
2251188bef
Fix ReaderLib.lib (and references)
2016-02-02 13:31:59 +01:00
f5d159bad6
Updating the timeline
2016-02-02 12:10:56 +00:00
52d6444920
Add new image reader under new reader architecture
2016-02-02 11:30:15 +01:00
f782ced23f
Fixing extensions
2016-02-01 14:32:01 +01:00
a22d3930ee
Renaming config
2016-02-01 13:50:59 +01:00
bc66cfc766
Limiting only to linux and release due to the licensing
2016-02-01 13:47:59 +01:00
172bfa2c3f
Changing the data directory
2016-02-01 13:47:59 +01:00
32c236c8c6
Adding baseline for linux
2016-02-01 13:47:59 +01:00
102302335a
fixing config typo
2016-02-01 13:47:59 +01:00
293e427dca
Fixing platform
2016-02-01 13:47:59 +01:00
4487745b5a
Removing ResNet for now
2016-02-01 13:47:59 +01:00
da0f233048
Updating the config + baselines
2016-02-01 13:47:59 +01:00
7818cadcbe
Adapting config to run in test environment
2016-02-01 13:47:59 +01:00
5e07d1c9f0
Adding resnet test
2016-02-01 13:47:59 +01:00
dcded35533
Adding AlexNet and ResNet directories
2016-02-01 13:47:59 +01:00
b0a6e00b8a
Use .cntk as default extension for CNTK configuration files
2016-02-01 10:19:48 +01:00
c4f9b262aa
Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/devicePlacement
2016-01-30 16:34:34 -08:00
931e72b19a
temporarily disabled Image tests in Jenkins test
2016-01-30 11:56:58 -08:00
b1e2753452
All image samples must take an imageLayout parameter
2016-01-30 11:56:57 -08:00
4baa3ce6de
(minor feedback from Vladimir)
2016-01-30 11:56:56 -08:00
a4f5b7a5d6
changed sample 02_Convolution to use imageLayout to enable cudnn use;
...
and fixed a refactoring bug in ImageParameter NDL definition
2016-01-30 11:56:55 -08:00
bee394dc0b
fixed an incorrect consistency check in TrainOrAdaptModel()
2016-01-30 11:56:54 -08:00
d2d8c010f0
changed example e2e tests to P leg
2016-01-30 11:56:53 -08:00
bd0b365213
reduced e2e time for ptb example from 11min to 6sec
2016-01-30 11:56:52 -08:00
7faf78ea9e
set data path linux vs. cygwin for MNIST e2e
2016-01-30 11:56:51 -08:00
16fe2c1a9c
chmod for run-test scripts
2016-01-30 11:56:51 -08:00
338318becf
Updated TestDriver script wrt gpu/cpu build under linux
2016-01-30 11:56:50 -08:00
ab765458bf
Changed yaml files for e2e example runs
2016-01-30 11:56:49 -08:00
d0e93543f7
Added examples as e2e tests
2016-01-30 11:56:48 -08:00
2b08c6a402
Simplification of GPU device selection code with removal of the parallel error-prone GetBestDevice functionality in GPUMatrix. Now device selection is fully centralized and AUTOPLACEMATRIX functionality is removed
2016-01-30 03:03:12 -08:00
dc1ed4628f
python 3 compatibility fix
2016-01-29 19:03:15 +01:00
73faaa7687
Remove <Scc*> tags from .vcxproj files
2016-01-29 11:21:06 +01:00
c7c5403994
tensor support for RowSlice operation;
...
completed CudaErrString<cublasStatus_t>()
2016-01-28 09:40:57 -08:00
4fe5fd7251
Modified projects and solution for a new CPUONLY configuration (Debug_CpuOnly and Release_CpuOnly)
...
Added missing CRT_SECURE_NO_WARNINGS to latticeNoGpu.cpp
Added missing reference to Matrix<double> (causing linking issues in Release_CpuOnly)
2016-01-27 16:16:37 +01:00
2462647f62
White-space (remove CRs that crept in)
2016-01-25 10:24:10 +01:00
af97079a58
removed #if-0'ed out code
2016-01-22 17:34:41 -08:00
a4cf44544a
normalized inconsistent commenting style from "//text" to "// text"
2016-01-22 13:58:47 -08:00
27641d86d0
normalized comment separator from /// to //
2016-01-22 13:47:40 -08:00
f4197414de
renamed EsotericNodes.h to SpecialPurposeNodes.h, and CompositeComputationNodes.h to PreComputeNodes.h;
...
moved GMMLogLikelihoodNode and SequenceWithSoftmaxNode to SpecialPurposeNodes.h;
also moved TimeReverseNode.h there (which will be deleted soon);
replaced the min/max macro use with std::min/max;
added all frequently used C/C++ headers into all stdafx.h
2016-01-22 10:45:53 -08:00
6f59eb1d22
Formatting
2016-01-22 09:23:03 +01:00
8e7b19b566
bug fix: LazyZeroGradient() must not use m_value dimensions for resizing m_gradient, because m_value may have been lost already due to memshare;
...
bug fix: DelayedValueBase should not call SetDims(., false) because upon re-load (rollback) there is already an MB Layout, which will cause a consistency check to fail
2016-01-21 11:57:03 -08:00
e389459ca6
LearnableParameter::InitRandom() now mimics the Matrix initialization behavior;
...
bug fix: LookupTableNode used GetAsMatrixNumRows() on input[1] which is a minibatch;
bug fix: Image/QuickE2E network definition updated to drop the now unnecessary extra column dimension;
bug fix: TensorShape::IsDense() should not require m_offset to be 0 (column slices are perfectly fine)
2016-01-20 23:28:33 -08:00
342b1ba460
changed the mapping of tensors onto the Matrix storage objects: In case of no MBLayout, the Matrix objects now always have column dimension 1. Actual matrices (as used by TimesNode) are now 2D tensors. As part of this, GetNumCols()/-Rows() no longer exist, but were split into GetSampleMatrixCols()/-Rows() (minibatch interpretation) and GetAsMatrixCols()/-Rows() (2D matrix interpretation)
2016-01-20 21:33:38 -08:00
78a7edeb36
BestGpu::Init() no longer throws when run under remote access
2016-01-18 16:40:31 -08:00
dcfc0e9679
License change
2016-01-18 09:36:17 +01:00
2f39fe3f01
Re-format code using clang-format (plus some post-processing)
2016-01-18 09:36:14 +01:00
1bcefe770a
.gitattributes: set proper configuration for text files (and normalize files)
2016-01-18 09:35:54 +01:00
eb0e08ecfd
Fixed testcases.yml to use build SKU gpu instead of default to align with the previous build SKU option value change in TestDriver
2016-01-15 14:31:32 -08:00
d5a3a89483
Changed default build location setting for Linux to accomodate multiple build SKUs
2016-01-15 10:44:52 -08:00
8ae5176a21
Merge branch 'master' of https://git01.codeplex.com/cntk into amitaga/separate1bitDataParallelSGD
2016-01-15 09:40:43 -08:00
a8029ea806
Changed TestDriver.py --build-sku option values
2016-01-15 09:40:00 -08:00
e9a91a8d38
Merge branch 'master' of https://git01.codeplex.com/cntk into bmitra/Changes
2016-01-15 03:30:27 -08:00
58d3987c43
Added missing parameter to matchesTag call in the run test command
2016-01-15 01:10:29 -08:00
252abe2de4
Added --build-sku option to TestDriver.py to enable filtering tests based on build SKUs (aka CPUONLY vs GPU vs 1BitSGD)
2016-01-15 00:08:07 -08:00
a1dc007413
Adding IsValid check and minor spell corrections in unit tests.
2016-01-14 09:02:48 -08:00
18528f15b4
Merge branch 'master' of https://git01.codeplex.com/cntk into amitaga/separate1bitDataParallelSGD
2016-01-13 22:43:27 -08:00
31b0e7098e
added a new file-format option to the MEL LoadModel command in order to load older CNTK files that had borked tensor dimensions;
...
FrameRange now allows time ranges
2016-01-13 18:42:17 -08:00
70cea38b08
Merge branch 'master' of https://git01.codeplex.com/cntk into amitaga/separate1bitDataParallelSGD
2016-01-13 09:58:40 -08:00
ed5d40aae1
changed GetRecurrenceDirections() back to operate on a singly dimension only (multiple dimensions can be realized with BrainScript);
...
towards implementing DataTensorFor() using tensor slices, so that we can reuse that for ShiftNode
2016-01-08 11:19:50 -08:00
08e6fc1b5c
towards multi-dimensional loops: new interface IRecurrentNode::GetRecurrenceDirections() implemented by old DelayedValueNode, ShiftNode, and partially by analysis code
2016-01-07 21:11:02 -08:00
41fb4273cc
fixed a compiler warning in Release (unused variable)
2016-01-07 18:10:07 -08:00
b770bdc73c
bug fix: incorrect Resize() in DefaultConvolutionEngine::Forward() destroyed content of matrix;
...
bug fix: last update of non-linearities must also correctly set OutputUsedInComputingInputNodesGradients() and likewise input;
bug fix: DelayedValueNode::ForwardProp() must correctly handle gaps also for first call before EndForwardProp();
beginnings of much extended DelayedValue node, called ShiftNode (not functional yet);
changed IStatefulNode::ImportState() take an rvalue reference (take ownership)
2016-01-07 17:06:33 -08:00
fe6e622614
Made inclusion of quantized gradient aggregation header compile-time conditional. Removed dependency on top-level MatrixQuantizer type from the quantizer unti tests - the tests now work with the underlying MatrixQuantizerImpl type.
2016-01-07 11:32:15 -08:00
6ed460bf96
Added a E2E test for buffered async gradient aggregation without quantization
2016-01-06 23:29:55 -08:00
8b03b5435f
Merge branch 'master' of https://git01.codeplex.com/cntk into amitaga/separate1bitDataParallelSGD
2016-01-06 18:39:06 -08:00
Gaizka Navarro
Frank Seide
Alexey Kamenev
Project Philly
Mark Hillebrand
Amit Agarwal
Philipp Kranen
Mark Hillebrand
Chris Basoglu
Eldar Akchurin
stonebig
bmitra