27f2af8 | Mark Hillebrand | 18 January 2016, 08:31:54 UTC | License change | 18 January 2016, 08:31:54 UTC |
d97a687 | Alexey Kamenev | 15 January 2016, 00:58:13 UTC | Updated ResNet ImageNet samples. | 15 January 2016, 00:58:13 UTC |
e4adfb7 | Alexey Kamenev | 14 January 2016, 20:53:41 UTC | Updated readme.txt | 14 January 2016, 20:53:41 UTC |
9d26271 | Alexey Kamenev | 14 January 2016, 18:49:33 UTC | Updated ResNet samples. | 14 January 2016, 20:32:48 UTC |
b3f7361 | Jasha Droppo | 08 January 2016, 22:34:29 UTC | TrainOrAdaptModel, when doing regularized adaptation, now properly INvalidateCompiledNetwork() before calling CompileNetwork() again. | 14 January 2016, 17:00:04 UTC |
f85ef91 | Jasha Droppo | 08 January 2016, 22:22:42 UTC | FromEvalOrder() warning message had possible nullptr dereference, fixed. | 14 January 2016, 17:00:03 UTC |
8d646e1 | Philipp Kranen | 14 January 2016, 10:12:20 UTC | removing zombie vcxproj files | 14 January 2016, 10:12:20 UTC |
135b992 | Philipp Kranen | 11 January 2016, 17:19:07 UTC | fixed Linux Makefile wrt deleted files | 14 January 2016, 08:06:40 UTC |
3eeb187 | Philipp Kranen | 11 January 2016, 13:18:13 UTC | reflected deleted files and example updates in VS solution | 14 January 2016, 08:06:40 UTC |
9e9cab0 | Philipp Kranen | 11 January 2016, 12:52:31 UTC | removing files and folders as announced (examples and readers) | 14 January 2016, 08:06:40 UTC |
fc61bb1 | Philipp Kranen | 18 December 2015, 14:28:51 UTC | Removed a few unreferencep source files | 14 January 2016, 08:03:10 UTC |
40cbeac | Frank Seide | 14 January 2016, 03:32:49 UTC | more strict control of compiled/non-compiled state of networks in MEL. All editing operations invalidate compilation, and Saving the model will validate the model if changed (if not it was validated during Load) | 14 January 2016, 03:32:49 UTC |
c89b632 | U-FAREAST\fseide | 14 January 2016, 02:56:33 UTC | Merge branch 'master' of https://git.codeplex.com/cntk into fseide/tensors | 14 January 2016, 02:56:33 UTC |
b3cfd6e | Frank Seide | 14 January 2016, 02:54:01 UTC | bug fix in TensorSliceWithMBLayoutFor() handling of m_timeOffset for entire batch | 14 January 2016, 02:54:01 UTC |
31b0e70 | Frank Seide | 14 January 2016, 02:42:17 UTC | 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 | 14 January 2016, 02:42:17 UTC |
de5be29 | thhoens | 13 January 2016, 00:23:04 UTC | Minor performance upgrade on row slicing to avoid GPU mem copy. | 13 January 2016, 23:45:20 UTC |
f835efd | thhoens | 13 January 2016, 00:01:29 UTC | Fix for multi GPU to share all parameters required to adjust learning rate. | 13 January 2016, 23:45:19 UTC |
d39d87f | thhoens | 09 January 2016, 12:24:23 UTC | Fixed a bug where the m_elemSizeAllocated was used instead of m_nz | 13 January 2016, 23:45:18 UTC |
55ccb09 | Yongqiang Wang | 13 January 2016, 22:16:12 UTC | Merge remote-tracking branch 'origin/master' into myseq | 13 January 2016, 22:16:12 UTC |
19a9895 | Yongqiang Wang | 13 January 2016, 20:25:20 UTC | Print SeqGammar related parameters for better logging. | 13 January 2016, 20:25:20 UTC |
ba61abd | Marko Radmilac | 09 January 2016, 00:41:45 UTC | Disable popups on Windows | 13 January 2016, 18:11:28 UTC |
270726e | Yongqiang Wang | 13 January 2016, 07:58:28 UTC | Make SequenceGammar calculation parameters configurable. | 13 January 2016, 07:58:28 UTC |
f0655f0 | Yongqiang Wang | 13 January 2016, 02:12:11 UTC | (make gcc happy) | 13 January 2016, 02:12:11 UTC |
c45401f | Yongqiang Wang | 13 January 2016, 01:08:46 UTC | Add Nesterov's momentum. to use NAG, simply add useNAG=true | 13 January 2016, 01:08:46 UTC |
b61d670 | Yongqiang Wang | 12 January 2016, 23:57:54 UTC | Merge remote-tracking branch 'origin/master' into myseq Conflicts: Source/CNTK/ModelEditLanguage.cpp Source/Common/Include/Sequences.h Source/ComputationNetworkLib/ComputationNetwork.h Source/ComputationNetworkLib/ComputationNetworkEditing.cpp Source/ComputationNetworkLib/ConvolutionalNodes.h | 12 January 2016, 23:57:54 UTC |
5bb9fbf | Yongqiang Wang | 12 January 2016, 23:50:02 UTC | Fix a bug pointed by Alexey. Thanks! | 12 January 2016, 23:50:02 UTC |
914ac61 | Yongqiang Wang | 12 January 2016, 23:27:32 UTC | Fix an inconsistency after merge with master. | 12 January 2016, 23:27:32 UTC |
7b0159a | Alexey Kamenev | 12 January 2016, 22:53:46 UTC | Added Python conversion script, updated readme.txt. | 12 January 2016, 22:53:46 UTC |
92e8a4d | Alexey Kamenev | 12 January 2016, 22:25:13 UTC | Added BN eval mode to MEL. Updated samples. | 12 January 2016, 22:47:32 UTC |
9e25b7e | Alexey Kamenev | 12 January 2016, 21:37:38 UTC | Removed Resize from BN code. Updated samples. | 12 January 2016, 22:02:51 UTC |
cc2a836 | Alexey Kamenev | 09 January 2016, 00:57:27 UTC | Updated samples, added ResNet-50. | 12 January 2016, 22:02:43 UTC |
f52e80c | Alexey Kamenev | 08 January 2016, 19:39:52 UTC | Added CMA to BN node, updated samples. | 12 January 2016, 22:02:37 UTC |
f764123 | Amit Agarwal | 12 January 2016, 06:45:17 UTC | Bug workaround: The m_columnsValidityMask matrix in MBLayout type was being default initialized resulting in incorrectly selecting a bad GPU device. | 12 January 2016, 06:45:17 UTC |
25fd18b | Yongqiang Wang | 12 January 2016, 01:26:52 UTC | Fix an error in SequenceWithSoftmaxNode::RequestMatricesBeforeForwardProp | 12 January 2016, 01:26:52 UTC |
314f057 | Yongqiang Wang | 12 January 2016, 01:07:56 UTC | Merge branch 'master' into myseq Conflicts: Source/ComputationNetworkLib/ComputationNode.h Source/ComputationNetworkLib/InputAndParamNodes.h Source/ComputationNetworkLib/TrainingCriterionNodes.h | 12 January 2016, 01:07:56 UTC |
2d7b74e | Yongqiang Wang | 22 December 2015, 20:18:21 UTC | (Fix a bug in MarkValueSharable) | 12 January 2016, 00:54:08 UTC |
8aa59f7 | Yongqiang Wang | 22 December 2015, 08:19:22 UTC | (further remove MarkValueNotSharable out of constructor) | 12 January 2016, 00:51:45 UTC |
60989d7 | Yongqiang Wang | 22 December 2015, 07:20:13 UTC | Move MarkValueNonsharable out of constuctors (make gcc happy) | 12 January 2016, 00:51:28 UTC |
6777bbe | Yongqiang Wang | 22 December 2015, 05:56:33 UTC | Display CUB and CUDNN paths (if defined) in BuildInfo Print BuildInfo at the very begining of the program. convenient for checking build type. | 12 January 2016, 00:49:00 UTC |
b54cfcc | Yongqiang Wang | 19 December 2015, 08:18:13 UTC | Add an alternate option "numSubminibatches" for users to indicate how to split minibatches into subminibatches. | 12 January 2016, 00:48:50 UTC |
2f51fb2 | Yongqiang Wang | 19 December 2015, 07:32:59 UTC | Fix a bug in MarkValueSharableNode | 12 January 2016, 00:48:40 UTC |
0b8e30e | Yongqiang Wang | 18 December 2015, 22:21:21 UTC | Revise the condition of ReleaseMatricesAfterForwardProp: only ValueSharable nodes can be released after forwardprop | 12 January 2016, 00:48:29 UTC |
4b1f800 | Yongqiang Wang | 18 December 2015, 01:44:22 UTC | Fix MarkValueNotSharableNodes | 12 January 2016, 00:48:13 UTC |
485a7b8 | Yongqiang Wang | 16 December 2015, 23:54:08 UTC | Revise the implementation of valueNotSharableNode. More to be revised. | 12 January 2016, 00:47:56 UTC |
edbb47d | Yongqiang Wang | 15 December 2015, 01:33:58 UTC | Replace CreateMatrixIfNull by MarkValueNonsharable() In the compiling the stage, we will mark nodes as nonsharable whose descendents are all learnable parameters. | 12 January 2016, 00:47:46 UTC |
569a4d6 | Yongqiang Wang | 12 January 2016, 00:32:19 UTC | Add support for revising batch normalization property in MEL. Now a BatchNormalization node's eval mode can be modified by SetProperty(BNnode, batchNormEvalMode, true); or by SetPropertyForSubTree(rootNode, batchNormEvalMode, true); in which all the BN nodes under rootNode will be changed. | 12 January 2016, 00:32:19 UTC |
bc3b96e | Yongqiang Wang | 11 January 2016, 20:29:22 UTC | Merge branch 'myseq' of //vilfblgpu011/d/users/erw/src/cntk into myseq | 11 January 2016, 20:29:22 UTC |
05e0262 | Yongqiang Wang | 10 January 2016, 07:17:51 UTC | Bug fix for ConvertDBN command | 10 January 2016, 07:17:51 UTC |
c41eafa | Yongqiang Wang | 10 January 2016, 07:16:45 UTC | Move all data member in MBLayout to CPU | 10 January 2016, 07:16:45 UTC |
4d9304c | Frank Seide | 10 January 2016, 05:34:44 UTC | removed multi-dim feature from ShiftNode, causes too many inconsistencies | 10 January 2016, 05:34:44 UTC |
08e7d59 | Frank Seide | 10 January 2016, 02:13:31 UTC | moved all actions (DoXXX()) from CNTK.cpp to ActionsLib (no code change); reenabled tensor lib (undid accidental commit) | 10 January 2016, 02:13:31 UTC |
69b5477 | Frank Seide | 10 January 2016, 01:03:20 UTC | (made sure non-tensor version still compiles) | 10 January 2016, 01:03:20 UTC |
0826c1c | Frank Seide | 10 January 2016, 00:58:26 UTC | moved non-tensor versions of PlusNode, MinusNode, and ElementTimesNode to EsoreticNodes.h (no code change) | 10 January 2016, 00:58:26 UTC |
a0fc021 | Frank Seide | 10 January 2016, 00:51:45 UTC | DataTensorFor() refactored further | 10 January 2016, 00:51:45 UTC |
c886f32 | Frank Seide | 10 January 2016, 00:30:25 UTC | factored out new function TensorSliceWithMBLayoutFor() from DataTensorFor(), for use by ShiftNode | 10 January 2016, 00:30:25 UTC |
5499741 | RuiZhao | 07 January 2016, 00:48:10 UTC | release after BP | 08 January 2016, 22:36:32 UTC |
a1427cc | RuiZhao | 29 December 2015, 23:00:09 UTC | release temp matrix in SE | 08 January 2016, 22:36:19 UTC |
1ad6c2d | RuiZhao | 24 December 2015, 21:48:21 UTC | frameskipv2 | 08 January 2016, 22:36:08 UTC |
ef84011 | RuiZhao | 23 December 2015, 19:39:39 UTC | SE frameskip V2 temp | 08 January 2016, 22:35:56 UTC |
fc33614 | RuiZhao | 18 December 2015, 23:17:48 UTC | frameskip SE | 08 January 2016, 22:35:36 UTC |
db9de92 | Frank Seide | 08 January 2016, 22:29:25 UTC | un-optimized DataTensorFor(), to again use the full tensor dimension (without pre-multiplying) | 08 January 2016, 22:29:25 UTC |
8f7f193 | Yongqiang Wang | 08 January 2016, 22:23:36 UTC | (make gcc happy) | 08 January 2016, 22:23:36 UTC |
ed5d40a | Frank Seide | 08 January 2016, 19:19:50 UTC | 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 | 08 January 2016, 19:19:50 UTC |
c1c818c | Frank Seide | 08 January 2016, 05:12:39 UTC | made gcc happy | 08 January 2016, 05:12:39 UTC |
08e6fc1 | Frank Seide | 08 January 2016, 05:11:02 UTC | towards multi-dimensional loops: new interface IRecurrentNode::GetRecurrenceDirections() implemented by old DelayedValueNode, ShiftNode, and partially by analysis code | 08 January 2016, 05:11:02 UTC |
a32704d | Frank Seide | 08 January 2016, 02:20:03 UTC | made gcc happy | 08 January 2016, 02:20:03 UTC |
7520910 | U-FAREAST\fseide | 08 January 2016, 02:11:30 UTC | Merge branch 'master' of https://git.codeplex.com/cntk into fseide/tensors | 08 January 2016, 02:11:30 UTC |
41fb427 | Frank Seide | 08 January 2016, 02:10:07 UTC | fixed a compiler warning in Release (unused variable) | 08 January 2016, 02:10:07 UTC |
6baed48 | Frank Seide | 08 January 2016, 01:13:27 UTC | made gcc happy | 08 January 2016, 01:13:27 UTC |
a961d71 | U-FAREAST\fseide | 08 January 2016, 01:07:06 UTC | Merge branch 'master' of https://git.codeplex.com/cntk into fseide/tensors | 08 January 2016, 01:07:06 UTC |
b770bdc | Frank Seide | 08 January 2016, 01:06:33 UTC | 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) | 08 January 2016, 01:06:33 UTC |
de6ac08 | Alexey Kamenev | 07 January 2016, 18:02:27 UTC | Fix for non-spatial BN dimensions. Updated samples. | 07 January 2016, 18:02:27 UTC |
dda0524 | Alexey Kamenev | 07 January 2016, 02:48:43 UTC | Fixed issue with MEL and convo engines, updated samples. | 07 January 2016, 02:50:28 UTC |
98d5b92 | Alexey Kamenev | 06 January 2016, 04:55:00 UTC | Updated samples. | 07 January 2016, 02:50:17 UTC |
71881dc | Alexey Kamenev | 06 January 2016, 01:33:02 UTC | Updated samples. | 07 January 2016, 02:50:07 UTC |
82dafa2 | Alexey Kamenev | 05 January 2016, 22:28:32 UTC | Fixed LearnableParameter and samples to work properly with cuDNN. | 07 January 2016, 02:49:56 UTC |
40ce1af | Amit Agarwal | 07 January 2016, 02:02:32 UTC | Added the ReleaseMatricesAfterBackprop function to sequence training node to release the temp matrices not needed after back propagation is complete for all the children of the nodes | 07 January 2016, 02:02:32 UTC |
87206fc | Amit Agarwal | 07 January 2016, 01:49:49 UTC | Fixed a corner case bug in async buffered gradient aggregation code | 07 January 2016, 01:49:49 UTC |
cb8cedd | Frank Seide | 06 January 2016, 23:11:02 UTC | changed non-linearity gradients to be computed from node output instead of node input (better for mem-sharing/in-place) | 06 January 2016, 23:11:02 UTC |
306e857 | Frank Seide | 06 January 2016, 16:18:32 UTC | renamed DataTensor.h to TensorShape.h. No code changes | 06 January 2016, 16:18:32 UTC |
4650ead | bmitra | 06 January 2016, 07:22:14 UTC | Merge branch 'master' of https://git01.codeplex.com/cntk into bmitra/Changes Conflicts: Source/ComputationNetworkLib/ReshapingNodes.h | 06 January 2016, 07:22:14 UTC |
aeb4a5a | Frank Seide | 06 January 2016, 04:36:24 UTC | accidentally disabled cudnn for Jenkins Image test | 06 January 2016, 04:36:24 UTC |
993f2ce | Frank Seide | 06 January 2016, 03:47:16 UTC | made gcc happy (wrong printf format) | 06 January 2016, 03:47:16 UTC |
51b7614 | Frank Seide | 06 January 2016, 03:42:08 UTC | added derived Reshape() variants ReshapeDimension(), FlattenDimensions(), and SplitDimension to BrainScript; undid accidental commit of a config file modified for a debug test | 06 January 2016, 03:42:08 UTC |
f8e7779 | Frank Seide | 06 January 2016, 03:21:03 UTC | implemented new ReshapeNode (which works with arbitrary tensor dims, but cannot swap dimensions into time--use TransposeNode for that in the future); old ReshapeNode renamed to DeprecatedReshapeNode. Still available from BrainScript for now, while new implementation is invoked as NewReshape(). This will change soon; bug fix: a few Validate() functions did not check isFinalValidationPass and failed on temporary input; bug fix: upwards-compatible Save/Load in convolution nodes by spltting one size_t into two uint32_t got the order wrong; bug fix: convolution nodes should only create convolution engines in final validation | 06 January 2016, 03:21:03 UTC |
4711562 | Frank Seide | 05 January 2016, 22:36:28 UTC | (fixed an error message) | 05 January 2016, 22:36:28 UTC |
30018b7 | Frank Seide | 05 January 2016, 22:19:16 UTC | added a new name "ImageParameter" for LearnableParameter that expects 3 input dimensions which are interpreted as WHC and mapped to actual tensor dimensions according to the optional imageLayout parameter. Adapted the NDL of the Image/QuickE2E test | 05 January 2016, 22:19:16 UTC |
2bf4d0a | bmitra | 05 January 2016, 20:50:35 UTC | Add missing definition to NoGPU.cpp. | 05 January 2016, 20:50:35 UTC |
7e780c3 | Frank Seide | 05 January 2016, 20:11:27 UTC | Past/FutureValue now takes a dimension tensor in BrainScript; removed cols/numImages parameter from InputValue and Past/FutureValue, since those always process data samples. Lots of little deletions in SimpleNetworkBuilder.cpp | 05 January 2016, 20:11:27 UTC |
d63576c | bmitra | 05 January 2016, 18:44:46 UTC | Minor fixes in convolution engine. | 05 January 2016, 18:44:46 UTC |
6f9b664 | Frank Seide | 04 January 2016, 22:20:38 UTC | (added comments) | 04 January 2016, 22:20:38 UTC |
c7dc6f1 | bmitra | 04 January 2016, 20:39:50 UTC | Adding comment for ReshapeNode::InferTargetSampleLayout() | 04 January 2016, 20:39:50 UTC |
e3757f0 | Yongqiang Wang | 04 January 2016, 20:26:53 UTC | A stopgap to prevent reader to load matrices inconsistent with lattices. Will be gone once the bug is fixed. | 04 January 2016, 20:26:53 UTC |
6899f5f | bmitra | 04 January 2016, 18:23:01 UTC | Merge branch 'master' of https://git01.codeplex.com/cntk into bmitra/Changes | 04 January 2016, 18:23:01 UTC |
ff9a916 | Frank Seide | 02 January 2016, 04:38:31 UTC | Coup de grâce: the 3 basic elementwise arithmetic operation (plus, minus, element times) and the 6 main non-linearity nodes (all except softmax and GMM) have been replaced by the new tensor library. This makes the code significantly more compact and allows broadcasting along any dimension, e.g. for implementing the image bias | 02 January 2016, 04:38:31 UTC |
5fc6f34 | Frank Seide | 02 January 2016, 03:20:35 UTC | bug fix in ConvolutionEngineTests.cpp: when testing CPU, use HWC layout | 02 January 2016, 03:20:35 UTC |
2752901 | Frank Seide | 02 January 2016, 03:14:10 UTC | bug fix: Pooling nodes initialized wrong conv-engine factory class | 02 January 2016, 03:14:10 UTC |
22e8ce3 | Frank Seide | 02 January 2016, 01:46:05 UTC | reenabled TensorView for PlusNode--nearly identical result Image/QuickE2E test | 02 January 2016, 01:46:05 UTC |
6d5df56 | Frank Seide | 02 January 2016, 01:31:45 UTC | temporarily disabled TensorView in PlusNode | 02 January 2016, 01:31:45 UTC |
730440e | Frank Seide | 02 January 2016, 01:27:36 UTC | added clipping of log and quotient, as SequenceTraining failed without | 02 January 2016, 01:27:36 UTC |