90a41f6 | Yuqing Tang | 27 March 2018, 23:48:11 UTC | Enabled gather op indcies be computed from parameters. | 27 March 2018, 23:48:11 UTC |
a7453ed | Thiago Crepaldi | 26 March 2018, 22:06:07 UTC | Merge pull request #3065 from Microsoft/szhao/fixsample Update cntk version to 2.5 in the sample | 26 March 2018, 22:06:07 UTC |
09cdc6c | Sheng Zhao (SPEECH) | 24 March 2018, 10:45:46 UTC | fix the cntk version in the sample | 24 March 2018, 10:45:46 UTC |
ed4bd3b | Project Philly | 23 March 2018, 01:33:54 UTC | Integrate jaliyaek/recur_checkin into master | 23 March 2018, 01:33:54 UTC |
a5e0912 | Thiago Crepaldi | 22 March 2018, 23:48:35 UTC | Fix automated cntk install script for windows | 23 March 2018, 00:16:50 UTC |
5b04f46 | Project Philly | 22 March 2018, 23:24:40 UTC | Integrate sptiwari/add_rnn_to_ornn2 into master | 22 March 2018, 23:24:40 UTC |
73c2046 | Jaliya Ekanayake | 22 March 2018, 18:44:09 UTC | Adding recurrence support to user defined functions. This enables UDF to be called inside recurrent loops. | 22 March 2018, 18:44:09 UTC |
3f72469 | Spandan Tiwari | 22 March 2018, 18:24:32 UTC | Adding ONNX support for rnnReLu and rnnTanh in OptimizedRNNStack. | 22 March 2018, 18:24:32 UTC |
65961c9 | Project Philly | 21 March 2018, 22:44:06 UTC | Integrate liqun/2954 into master | 21 March 2018, 22:44:06 UTC |
82142f0 | Project Philly | 21 March 2018, 21:19:55 UTC | Integrate muditj/FixNoRandomizerBug3 into master | 21 March 2018, 21:19:55 UTC |
aab2567 | Project Philly | 21 March 2018, 19:23:28 UTC | Integrate liqun/RNNStage into master | 21 March 2018, 19:23:28 UTC |
c8c02bf | Mudit Jain | 21 March 2018, 18:57:19 UTC | Fix no randomizer bug where workers see duplicate data | 21 March 2018, 18:57:19 UTC |
3e83c56 | liqfu | 21 March 2018, 18:14:28 UTC | issue 2954: c# API memory leak | 21 March 2018, 18:14:28 UTC |
08ba969 | liqfu | 21 March 2018, 16:57:54 UTC | support ONNX RNN | 21 March 2018, 16:57:54 UTC |
3a22962 | Project Philly | 21 March 2018, 05:23:48 UTC | Integrate sptiwari/mean_var_norm_issue into master | 21 March 2018, 05:23:48 UTC |
92b7b27 | KeDengMS | 21 March 2018, 01:57:47 UTC | Update .clang-format to remove column limits Otherwise too much code change when formatting with clang-format | 21 March 2018, 01:57:47 UTC |
7068329 | Spandan Tiwari | 20 March 2018, 17:53:06 UTC | Fixes to mean_variance_normalization and ONNX LSTM to run Office models. | 20 March 2018, 17:53:06 UTC |
10c327e | Junjie Qian | 14 March 2018, 18:39:44 UTC | No aggregation or NCCL initialization if only one node used undef _SCL_SECURE_NO_WARNINGS for training | 19 March 2018, 16:54:41 UTC |
542406d | Junjie Qian | 13 March 2018, 22:21:59 UTC | No need to aggregate with one worker in brainscript Remove msg printint GPU UUID | 19 March 2018, 16:50:09 UTC |
91dbed4 | Project Philly | 17 March 2018, 20:32:53 UTC | Integrate vadimma/readse into master | 17 March 2018, 20:32:53 UTC |
6942ad1 | Thiago Crepaldi | 17 March 2018, 02:59:16 UTC | Update README.md with correct links to examples | 17 March 2018, 02:59:16 UTC |
7e1a653 | KeDengMS | 17 March 2018, 00:46:46 UTC | Update README and .clangformat | 17 March 2018, 00:46:46 UTC |
296ac76 | Thiago Crepaldi | 15 March 2018, 05:30:53 UTC | Preparing Release 2.5 - update documentation | 16 March 2018, 22:21:50 UTC |
34ed567 | Thiago Crepaldi | 15 March 2018, 04:19:01 UTC | Preparing Release 2.5 - bump version number | 16 March 2018, 22:21:38 UTC |
e39aeae | Project Philly | 15 March 2018, 17:03:41 UTC | Integrate liqun/it20183 into master | 15 March 2018, 17:03:41 UTC |
0f428b3 | liqfu | 15 March 2018, 15:47:22 UTC | iteration update | 15 March 2018, 15:47:22 UTC |
0b79085 | Spandan Tiwari | 15 March 2018, 07:21:47 UTC | Fixing ONNX producer_version bug. | 15 March 2018, 07:21:47 UTC |
4178d9f | Project Philly | 15 March 2018, 04:09:25 UTC | Integrate liqun/ONNXIRupdateStage2 into master | 15 March 2018, 04:09:25 UTC |
7fd48e7 | Spandan Tiwari | 15 March 2018, 00:22:54 UTC | Updating release notes with ONNX and other feature work. | 15 March 2018, 00:22:54 UTC |
a895b76 | liqfu | 14 March 2018, 22:28:08 UTC | update with ONNXIR | 14 March 2018, 22:28:08 UTC |
3a1b8e1 | liqfu | 14 March 2018, 21:47:58 UTC | update with ONNXIR | 14 March 2018, 21:47:58 UTC |
18723c7 | Project Philly | 14 March 2018, 19:00:05 UTC | Integrate liqun/GRUStage into master | 14 March 2018, 19:00:05 UTC |
6608da7 | Vadim Mazalov | 10 March 2018, 02:44:19 UTC | Add logs for reader debugging | 14 March 2018, 18:25:20 UTC |
46b6016 | liqfu | 14 March 2018, 16:30:30 UTC | Add ONNX GRU support | 14 March 2018, 16:30:30 UTC |
feb00c1 | Project Philly | 14 March 2018, 02:30:11 UTC | Integrate yuqtang/ValidateRecurerenceStepFunction into master | 14 March 2018, 02:30:11 UTC |
e61a7f0 | Jaliya Ekanayake | 13 March 2018, 23:51:06 UTC | Adding distributed GAN example | 13 March 2018, 23:51:06 UTC |
edbda9d | Yuqing Tang | 13 March 2018, 21:38:57 UTC | Added sanitary check for step function signature with appropriate error messages. | 13 March 2018, 21:38:57 UTC |
f1f3bb4 | liqfu | 13 March 2018, 18:24:33 UTC | Support ONNX GRU | 13 March 2018, 18:24:33 UTC |
44c626a | Thiago Crepaldi | 12 March 2018, 21:11:31 UTC | Fix wheel package name for nightly build scripts WHL package name has recently changed for GPU target build. Now the package is called cntk-gpu, which translates to a filename prefixed as cntk_gpu*.whl | 13 March 2018, 17:03:28 UTC |
b3daad9 | Thiago Crepaldi | 07 March 2018, 23:57:47 UTC | Add 'long_description' for wheel package's metadata file | 13 March 2018, 17:02:26 UTC |
769ea5c | Spandan Tiwari | 13 March 2018, 04:06:04 UTC | Adding OptimizedRNNStack ONNX support. Also Identity and NoOp support. | 13 March 2018, 04:06:04 UTC |
4b5c085 | Jaliya Ekanayake | 12 March 2018, 15:26:56 UTC | Fix the default device issue in test | 12 March 2018, 15:26:56 UTC |
82b56d1 | Jaliya Ekanayake | 12 March 2018, 08:00:49 UTC | Fixing the distributed learner sample count issue | 12 March 2018, 08:00:49 UTC |
ae81129 | Project Philly | 10 March 2018, 03:27:56 UTC | Integrate jqian/nccl into master | 10 March 2018, 03:27:56 UTC |
3d0af05 | Project Philly | 10 March 2018, 02:16:33 UTC | Integrate kedeng/fixCSBaseline into master | 10 March 2018, 02:16:33 UTC |
00c7944 | KeDengMS | 09 March 2018, 23:37:05 UTC | Fix OneHotSkip for C# 32-bit SizeTVector | 09 March 2018, 23:37:05 UTC |
405f4be | Project Philly | 09 March 2018, 23:27:46 UTC | Integrate kedeng/fix3004B into master | 09 March 2018, 23:27:46 UTC |
966f7ad | Thiago Crepaldi | 07 March 2018, 23:57:47 UTC | Update pip package metadata (setup.py) | 09 March 2018, 22:00:55 UTC |
d63ddd9 | KeDengMS | 09 March 2018, 19:48:53 UTC | Fix Issue #3004 for mpi=no | 09 March 2018, 19:48:53 UTC |
487a8bb | KeDengMS | 09 March 2018, 19:18:59 UTC | Update baseline following commit 678b8a9 | 09 March 2018, 19:18:59 UTC |
b709af3 | Project Philly | 09 March 2018, 08:32:30 UTC | Integrate kedeng/fix3019B into master | 09 March 2018, 08:32:30 UTC |
3cb3430 | Project Philly | 09 March 2018, 01:45:36 UTC | Integrate 678b8a9a986b4423449fca4ecdd2a492438dcc28 into master | 09 March 2018, 01:45:36 UTC |
f6b3260 | KeDengMS | 09 March 2018, 00:37:09 UTC | Fix issue #3019: RuntimeError: OptimizedRNNStackNode configured for sequence mode, but minibatch only has one time step. | 09 March 2018, 00:37:09 UTC |
28f587f | Junjie Qian | 02 March 2018, 20:10:06 UTC | Get GPU ID with nvmlDeviceGetUUID instead of device id in /proc/ Thanks to Frank's initial code change. Initialize NCCL with GPU ID frm nvml call instead of from device id in /proc. Previous way causes "same device being used" error if two different containers running on same host and try to initialize NCCL. Add nvml library for nvml apis Update the Makefile | 08 March 2018, 18:41:37 UTC |
dd88885 | Junjie Qian | 02 March 2018, 05:11:57 UTC | Fall back to MPI if failed to initialize NCCL | 08 March 2018, 18:41:37 UTC |
4f03bad | KeDengMS | 08 March 2018, 07:11:22 UTC | Fix memory leak in MKL layout | 08 March 2018, 07:11:22 UTC |
7215678 | KeDengMS | 08 March 2018, 03:13:17 UTC | Increase MPI pack threshold for more efficient distributed training Expose the knob to change MPI pack threshold in V2 Fixed a build error with mpi=no | 08 March 2018, 03:13:17 UTC |
85c1429 | Thiago Crepaldi | 01 March 2018, 17:32:45 UTC | Rename ImageWriterDll to DelayLoadedExtensionsDll | 08 March 2018, 01:15:04 UTC |
cb1a23b | mx-iao | 08 March 2018, 00:34:57 UTC | Merge pull request #2951 from bigonazzi/patch-1 Fix typo | 08 March 2018, 00:34:57 UTC |
b8ec6a7 | mx-iao | 08 March 2018, 00:34:15 UTC | Merge pull request #2959 from tcleberg/master removed dead link easons to switch to CNTK from Tensorflow from read… | 08 March 2018, 00:34:15 UTC |
4144407 | mx-iao | 08 March 2018, 00:33:52 UTC | Merge pull request #2978 from gabrielilharco/master Fixing typo in Manual_How_to_feed_data.ipynb | 08 March 2018, 00:33:52 UTC |
7dc7b53 | mx-iao | 07 March 2018, 23:38:42 UTC | Create ThirdPartyNotices.md | 07 March 2018, 23:38:42 UTC |
0f28edf | Thiago Crepaldi | 06 March 2018, 21:17:20 UTC | Fix CNTK build and tests after 1bitSGD merge | 07 March 2018, 23:31:25 UTC |
819baf0 | mx-iao | 07 March 2018, 23:27:17 UTC | Update LICENSE.md | 07 March 2018, 23:27:17 UTC |
678b8a9 | Nikos Karampatziakis | 07 March 2018, 04:42:51 UTC | Update cntk_cs.i | 07 March 2018, 04:42:51 UTC |
1d03a83 | KeDengMS | 06 March 2018, 23:21:31 UTC | Simplify h-softmax interface | 06 March 2018, 23:21:31 UTC |
20056a5 | Thiago Crepaldi | 06 March 2018, 18:59:34 UTC | Merge remote-tracking branch '1BitSGD/master' into thiagofc/master | 06 March 2018, 18:59:34 UTC |
acc72dc | Project Philly | 06 March 2018, 18:21:24 UTC | Integrate liqun/lstmstage2 into master | 06 March 2018, 18:21:24 UTC |
06f32e8 | Thiago Crepaldi | 06 March 2018, 00:07:39 UTC | Removing 1BitSGD submodule prior to merge with 1BitSGD repo | 06 March 2018, 00:09:51 UTC |
8eb1597 | Thiago Crepaldi | 05 March 2018, 21:56:51 UTC | Reorganizing 1BitSGD files prior to merge with CNTK. See details below Custom license was removed Headers were updated to reflect MIT license README.md and .gitattributes were removed (CNTK repo already have them) | 05 March 2018, 23:02:53 UTC |
7651b0d | liqfu | 05 March 2018, 21:13:03 UTC | ONNX LSTM support | 05 March 2018, 21:13:03 UTC |
6a8773b | Mark Hamilton | 01 March 2018, 19:17:13 UTC | Bump pom.xml version | 04 March 2018, 21:23:49 UTC |
de52f50 | Mark Hamilton | 28 February 2018, 22:02:37 UTC | Update pom.xml | 04 March 2018, 21:23:49 UTC |
0b118c8 | Ratan Rai Sur | 26 July 2017, 18:48:45 UTC | Enable scripted maven release rename maven-deploy | 04 March 2018, 21:23:49 UTC |
21a71fc | Ratan Rai Sur | 25 July 2017, 14:58:30 UTC | update groupID to match maven groupid add more maven deploy changes | 04 March 2018, 21:23:49 UTC |
35ce24a | Ratan Rai Sur | 25 July 2017, 14:58:10 UTC | update java make to make sources and javadoc jars | 04 March 2018, 21:23:49 UTC |
7d8e135 | Ratan Rai Sur | 24 July 2017, 17:18:02 UTC | add pom.xml for maven central publishing | 04 March 2018, 21:23:49 UTC |
677cb42 | Spandan Tiwari | 03 March 2018, 06:37:13 UTC | Updated MeanVarianceNormalization and LayerNormalization to work with epsilon. | 03 March 2018, 06:37:13 UTC |
311110d | Project Philly | 02 March 2018, 23:51:13 UTC | Integrate kedeng/fixShapeInfer2 into master | 02 March 2018, 23:51:13 UTC |
cc30e2e | Jaliya Ekanayake | 02 March 2018, 02:27:34 UTC | Adding distributed learner support for distributed training | 02 March 2018, 02:27:34 UTC |
d770419 | KeDengMS | 01 March 2018, 22:47:46 UTC | Fix shape inference in step function for scalar to broadcast | 01 March 2018, 22:47:46 UTC |
9af853d | Vadim Mazalov | 01 March 2018, 03:42:11 UTC | Ensure we can read model from a different location | 01 March 2018, 16:10:39 UTC |
cc163da | Manik Jindal | 28 February 2018, 23:08:19 UTC | Adding nightly builds links in news Adding nightly builds links in news for public visibility. | 28 February 2018, 23:08:19 UTC |
6d487c0 | Manik Jindal | 28 February 2018, 22:43:01 UTC | Add links to nightly release pages | 28 February 2018, 22:43:01 UTC |
4506eff | Manik Jindal | 28 February 2018, 00:26:05 UTC | Add support to download WHL packages for RC builds | 28 February 2018, 00:26:05 UTC |
9c5dba4 | Project Philly | 27 February 2018, 23:47:04 UTC | Integrate jaliyaek/bugfix_sq into master | 27 February 2018, 23:47:04 UTC |
bd74694 | Jaliya Ekanayake | 27 February 2018, 19:43:53 UTC | Fix the converter bug that omits cloning parts of the graph | 27 February 2018, 19:43:53 UTC |
772f0d9 | Thiago Crepaldi | 14 February 2018, 22:49:32 UTC | Remove 3rd party libs and debug info wheel packages | 27 February 2018, 17:29:16 UTC |
69850fd | Project Philly | 27 February 2018, 02:29:55 UTC | Integrate 0b31b08f9edd4ee8aa35445b8084a86d9c421419 into master | 27 February 2018, 02:29:55 UTC |
d1925a1 | Mark Hamilton | 23 February 2018, 18:04:24 UTC | Fix dependency loading in java | 26 February 2018, 20:28:20 UTC |
1d676a0 | Mark Hamilton | 20 February 2018, 21:29:09 UTC | Add dependencies to java jar | 26 February 2018, 20:28:20 UTC |
89ba54e | mx-iao | 26 February 2018, 18:31:18 UTC | Updating License to "Cognitive Toolkit" from "Computational Network Toolkit" | 26 February 2018, 18:31:18 UTC |
0b31b08 | Nikolay Amiantov | 25 February 2018, 18:02:46 UTC | Include <functional> in EvaluatorWrapper This fixes compilation failure. | 25 February 2018, 18:03:21 UTC |
695bdf7 | KeDengMS | 23 February 2018, 19:21:23 UTC | Fix crash in CBF when crossing sweep boundary This fixes issue #2479 #2964 | 23 February 2018, 19:21:23 UTC |
523af50 | Vadim Mazalov | 21 February 2018, 19:14:22 UTC | Minor refactoring | 21 February 2018, 19:16:19 UTC |
1c78107 | Vadim Mazalov | 17 February 2018, 02:21:52 UTC | Handle malformed lattice | 21 February 2018, 19:16:19 UTC |
9032174 | KeDengMS | 21 February 2018, 00:18:38 UTC | Disable MKL-DNN for now before we pick up the fix for AMD cache size fix. In MKL-DNN 0.12 release, there's a cache size detection issue causing crash on AMD Ryzen CPUs. This change disables MKL-DNN to unblock AMD Ryzen users for now. | 21 February 2018, 00:18:38 UTC |
461b82d | Manik Jindal | 20 February 2018, 18:54:07 UTC | Move hard-coded CNTK version to a common place This will help to build CNTK for nighlty builds with required CNTK version. If environment variable 'BUILD_CNTK_VERSION' is set, then CNTK will be build for that version as a public release('+' won't be appended to CNTK version). Otherwise hard-coded CNTK version will be used as a private build (2.4+). | 20 February 2018, 21:22:59 UTC |
84934ed | Gabriel Ilharco Magalhaes | 20 February 2018, 19:27:11 UTC | Fixing typo | 20 February 2018, 19:27:11 UTC |
4017a16 | Spandan Tiwari | 16 February 2018, 08:41:08 UTC | Adding mean_variance_normalization CNTK and ONNX op, and LayerNormalization ONNX suort. | 16 February 2018, 08:41:08 UTC |