f2a6ccb | marhamil723 | 12 October 2017, 14:40:03 UTC | rename tmpdir | 12 October 2017, 14:40:03 UTC |
0e4fc62 | Ratan Rai Sur | 11 October 2017, 20:19:05 UTC | use CNTKNativeUtils load in java | 11 October 2017, 20:19:05 UTC |
c37d163 | Ratan Rai Sur | 07 August 2017, 18:18:17 UTC | make java so variable names in makefile less confusing | 11 October 2017, 14:25:00 UTC |
5ee9c37 | marhamil723 | 20 July 2017, 14:59:03 UTC | cudnn64_5 -> cudnn64_6 | 11 October 2017, 14:25:00 UTC |
ce20931 | marhamil723 | 19 July 2017, 19:36:59 UTC | add cuda deps | 11 October 2017, 14:25:00 UTC |
844ca58 | marhamil723 | 18 July 2017, 15:24:23 UTC | fix sln, and make post-build give better message | 11 October 2017, 14:25:00 UTC |
028175f | Ratan Rai Sur | 17 July 2017, 15:45:47 UTC | use correct mkl lib path for jar | 11 October 2017, 14:25:00 UTC |
83c984f | marhamil723 | 19 May 2017, 19:30:08 UTC | add DLL's and so's to jar (squashed) minor fixes fixing Native utils path, loader, and the Native Load Utils remove path and classpath setting in tests cleaning up native utils Unified CNTKNativeUtils class. Changed the code generation to use CNTKNativeUtils directly instead of the intermediary CNTK.init(). adding fixes to post bild Added NATIVE_LOAD_MANIFEST for the cases when only specific high-level libraries should be loaded linux side Add gpu support linux gpu .so copying | 11 October 2017, 14:24:54 UTC |
23878e5 | Mark Hillebrand | 15 September 2017, 07:23:14 UTC | git checkout 741d01e36b1343d987dc85cfc8d7964816de97a8 Documentation/current_iteration.md | 15 September 2017, 07:23:14 UTC |
3a3a6a1 | Yuqing Tang | 14 September 2017, 23:23:44 UTC | Squashed commit of the following: Updated the manual on how to use learners to explain the new Learner APIs for release 2.2. | 15 September 2017, 06:21:29 UTC |
06a06c3 | Yuqing Tang | 14 September 2017, 22:19:29 UTC | Corrected the case of the word "version". | 15 September 2017, 06:14:45 UTC |
9613e7c | Yuqing Tang | 14 September 2017, 22:12:17 UTC | Squashed commit of the following: Removed sensitive information in CNTK start-up message. | 15 September 2017, 06:14:27 UTC |
14cd336 | Yuqing Tang | 14 September 2017, 20:01:55 UTC | Squashed commit of the following: Add python side work-around to allow the combination of mean gradient and specifying number samples with the right implementation; in Python, achieve the legacy logic of having per sample learning rate with mean gradient by overriding the per sample learning rate with ignored_minibatch_size. Also clarification in docstring. | 15 September 2017, 06:14:08 UTC |
e6e15ee | Spandan Tiwari | 13 September 2017, 20:34:20 UTC | Minor updates to FastRCNN readme files. | 15 September 2017, 05:57:40 UTC |
a7f93cc | Mark Hillebrand | 13 September 2017, 14:20:33 UTC | Tests/Install/linux/prep-run-test.sh: do not run CNTK_104_Finance_Timeseries_Basic_with_Pandas_Numpy.ipynb | 13 September 2017, 14:20:33 UTC |
b69b34e | liqun fu | 12 September 2017, 18:31:34 UTC | correct CSharp name, disable a flaky test for further investigation | 13 September 2017, 00:22:07 UTC |
b9f4472 | hecli | 12 September 2017, 20:22:38 UTC | CNTK 2.2 | 12 September 2017, 20:22:38 UTC |
08c0e47 | hecli | 12 September 2017, 20:08:40 UTC | Bump version numbers | 12 September 2017, 20:08:40 UTC |
7c283e5 | Mark Hillebrand | 23 May 2017, 09:32:43 UTC | Rename Tests/EndToEndTests/CNTKv2Python/Examples/{deeprl,DeepQNeuralNetwork}_test{ -> _disable}.py. Not for integration. | 12 September 2017, 18:28:37 UTC |
5cc1af9 | Mark Hillebrand | 20 May 2017, 13:26:51 UTC | Scripts/install/linux/conda-*.yml: gym[atari] -> gym So as to not require gcc for binary drop installation. | 12 September 2017, 18:28:24 UTC |
010903e | Zhou Wang | 29 August 2017, 12:44:47 UTC | remove inclusion of dependency dlls, as it has been referenced in UWP nuget package. | 12 September 2017, 18:07:54 UTC |
f004aea | Zhou Wang | 12 September 2017, 16:46:33 UTC | add information about examples to relesae notes | 12 September 2017, 16:59:18 UTC |
bfce957 | Project Philly | 12 September 2017, 15:54:59 UTC | Integrate liqun/LogisticRegFix into master | 12 September 2017, 15:54:59 UTC |
351e702 | Mark Hillebrand | 12 September 2017, 11:18:10 UTC | Tests/Install/linux/prep-run-test.sh: bump Jupyter TO to 2700 | 12 September 2017, 14:57:58 UTC |
e5b4bdc | liqun fu | 12 September 2017, 14:43:36 UTC | quick fix logistic regression build break | 12 September 2017, 14:43:36 UTC |
c76b6f0 | Mark Hillebrand | 12 September 2017, 12:58:09 UTC | Tests/EndToEndTests/TestDriver.py: add major CC for Tesla M60 | 12 September 2017, 12:58:09 UTC |
f89f9d7 | Mark Hillebrand | 12 September 2017, 08:45:21 UTC | Scripts/install/linux/install-cntk.sh: install libtiff5 | 12 September 2017, 10:17:45 UTC |
ab8f4e5 | Project Philly | 12 September 2017, 08:10:48 UTC | Integrate liqun/MakeDrop2.2Stage into master | 12 September 2017, 08:10:48 UTC |
0f4044e | Project Philly | 12 September 2017, 07:27:40 UTC | Integrate nikosk/attention4 into master | 12 September 2017, 07:27:40 UTC |
f41a1d2 | Project Philly | 12 September 2017, 05:44:31 UTC | Integrate sptiwari/remove_installfastrcnn into master | 12 September 2017, 05:44:31 UTC |
16d01d7 | Project Philly | 12 September 2017, 05:35:08 UTC | Integrate yuqtang/StagingRefactorLearningCore into master | 12 September 2017, 05:35:08 UTC |
738b740 | Nikos Karampatziakis | 12 September 2017, 02:43:46 UTC | Attention overhaul | 12 September 2017, 02:43:46 UTC |
10a68a4 | Project Philly | 12 September 2017, 01:57:48 UTC | Integrate liqun/CSApiRegressionExample into master | 12 September 2017, 01:57:48 UTC |
2b3e104 | Project Philly | 12 September 2017, 01:50:31 UTC | Integrate DaveyBiggers-master into master | 12 September 2017, 01:50:31 UTC |
9c34c47 | Project Philly | 12 September 2017, 01:45:07 UTC | Integrate liqun/CSApiRegressionExampleStage into master | 12 September 2017, 01:45:07 UTC |
ccfd6c7 | Yuqing Tang | 12 September 2017, 00:36:39 UTC | Refactor the core learner APIs: * Simplified TrainingParameterSchedule * Added minibatch_size to TrainingParameterSchedule and Learner * Added minibatch size to Learner * Added learning_parameter_schedule to Python APIs * Marked as deprecated on UnitType, learning_parameter_schedule, momentum_as_time_constant_schedule * Added minibatch_size parameter to all the learners: sgd, momentum_sgd, adam and all others except for user defined learner * Simplified learner parameter logic: If minibatch_size == cntk.learners.IGNORE, act as the parameters are applied applied to the whole minibatch; otherwise, scaling the parameters according to the actual minibatch_size and the minibatch_size specified during the the schedules and the learners construction time. More subtle details are in the docstrings and the comments. | 12 September 2017, 00:36:39 UTC |
e70a023 | Project Philly | 11 September 2017, 23:33:03 UTC | Integrate chenta/free_dimension_pooling into master | 11 September 2017, 23:33:03 UTC |
a3f20af | liqfu | 11 September 2017, 19:51:47 UTC | Add PretrainedModels to windows/linux/examples drops | 11 September 2017, 19:51:47 UTC |
86315fc | liqun fu | 11 September 2017, 19:39:12 UTC | Add a Hello World Logistic Regression Example | 11 September 2017, 19:39:12 UTC |
b84e10a | Spandan Tiwari | 11 September 2017, 19:03:46 UTC | Removing vestigial references to install_fastrcnn.py. | 11 September 2017, 19:03:46 UTC |
f07a05d | Cheng Tang | 08 September 2017, 18:47:07 UTC | remove check for free dimension on pooling add ut add test with strides input | 11 September 2017, 18:00:02 UTC |
6aef302 | Spandan Tiwari | 11 September 2017, 17:21:46 UTC | Enabling free static axes support for reshape API syntaxes. | 11 September 2017, 17:21:46 UTC |
ab7b4ff | Emad Barsoum | 11 September 2017, 16:23:33 UTC | Merge branch 'master' of https://github.com/DaveyBiggers/CNTK into DaveyBiggers-master Pull Request. | 11 September 2017, 16:23:33 UTC |
f52e724 | Mark Hillebrand | 11 September 2017, 11:07:50 UTC | Source/CNTKv2LibraryDll/tensorboard/TensorBoardUtils.cpp: ensure libopencv_imgproc is picked up as direct link dependency on Linux | 11 September 2017, 11:23:33 UTC |
0c54a7c | Emad Barsoum | 09 September 2017, 05:54:44 UTC | Add reduction rank to Convolution and ConvolutionTranspose at the C++ API and low level python API. | 09 September 2017, 05:54:44 UTC |
19b74f5 | liqun fu | 09 September 2017, 01:46:40 UTC | Add Logistic Regression Example | 09 September 2017, 01:46:40 UTC |
7a03064 | REDMOND\sayanpa | 08 September 2017, 20:43:38 UTC | Re-enable disabled doc test and address a pending CR comment in tutorial documentation. | 08 September 2017, 20:43:38 UTC |
cc788b8 | Project Philly | 08 September 2017, 19:03:00 UTC | Integrate jqian/wip/nccl2.0 into master | 08 September 2017, 19:03:00 UTC |
1cd345a | Spandan Tiwari | 08 September 2017, 18:41:56 UTC | Minor formatting fix to FastRCNN Readme.md. | 08 September 2017, 18:41:56 UTC |
9e6055b | Spandan Tiwari | 08 September 2017, 18:40:54 UTC | Updated FastRCNN Readme.md with new Linux Cython module info and other minor edits. | 08 September 2017, 18:40:54 UTC |
d820eba | Spandan Tiwari | 08 September 2017, 18:36:52 UTC | Updated Readme.md with new Linux Cython modules info and minor text edits. | 08 September 2017, 18:36:52 UTC |
37c3b40 | Junjie Qian | 08 September 2017, 17:31:13 UTC | CR comments | 08 September 2017, 17:31:28 UTC |
d899999 | Junjie Qian | 26 April 2017, 00:32:33 UTC | Change for NCCL 2.0 Fix race condition in Nccl init that may cause hang in NCCL2 | 08 September 2017, 17:31:28 UTC |
558d80c | Project Philly | 08 September 2017, 14:47:47 UTC | Integrate zhouwang/allow-freeaxis-master into master | 08 September 2017, 14:47:47 UTC |
bd048f6 | Ivan Rodriguez | 28 August 2017, 16:04:14 UTC | Uncommenting lines | 08 September 2017, 12:33:27 UTC |
73d798a | Zhou Wang | 08 September 2017, 06:21:20 UTC | improve error message; add tests for CopyVariableValueTo with free dimensions update baseline adjust tests to make sure batch length > 1 in case this is required. | 08 September 2017, 12:18:21 UTC |
f00507e | Zhou Wang | 07 September 2017, 18:57:45 UTC | Make sure that inferredVarShape has correct rank in order to avoid any rank resize during inferring free dimension. | 08 September 2017, 12:18:20 UTC |
073bf1d | Project Philly | 08 September 2017, 12:12:57 UTC | Integrate t-ivrodr/adding_more_cpp_cs_eval_examples into master | 08 September 2017, 12:12:57 UTC |
6346abe | Ivan Rodriguez | 28 August 2017, 16:04:14 UTC | Addressing comments in CR | 08 September 2017, 09:50:27 UTC |
f247772 | REDMOND\sayanpa | 07 September 2017, 23:33:26 UTC | Enabled conversion of jupyter notebook to html using nbsphinx, integrated the conversion with the current doc build system, updated the jupyter notebooks in Tutorial and Manual folders to adhere to nbshpinx guidelines | 07 September 2017, 23:33:26 UTC |
1cca6e2 | Project Philly | 07 September 2017, 23:06:10 UTC | Integrate duli/tensorboard_image into master | 07 September 2017, 23:06:10 UTC |
15953bf | Project Philly | 07 September 2017, 21:40:52 UTC | Integrate kedeng/sparseAgg2 into master | 07 September 2017, 21:40:52 UTC |
1d5937e | Project Philly | 07 September 2017, 19:09:59 UTC | Integrate liqun/CSApiFix into master | 07 September 2017, 19:09:59 UTC |
7d66c47 | KeDengMS | 07 September 2017, 06:18:49 UTC | Sparse gradient aggregation on GPU This change saves costly conversion from sparse to dense before gradient aggregation when embedding vocabulary size is huge. It is currently enabled for GPU build when training on GPU with non-quantized data parallel SGD. For other distributed learners and CPU build, it is disabled by default. It can be manually turned off in python by calling `cntk.cntk_py.use_sparse_gradient_aggregation_in_data_parallel_sgd(False)` Note that for a rare case of running distributed training with CPU device on a GPU build, you need to manually turn it off to avoid unimplemented exception | 07 September 2017, 06:18:49 UTC |
ee04a89 | duli1 | 07 September 2017, 05:35:17 UTC | Tensorboard image feature support. | 07 September 2017, 05:35:17 UTC |
795380a | liqun fu | 07 September 2017, 04:57:09 UTC | fix incorrect references/nuget version/typeos. Update CSTraining readme.md | 07 September 2017, 04:57:09 UTC |
765707a | Spandan Tiwari | 06 September 2017, 23:48:41 UTC | Make convolution work with free static axes coming from sequence unpack. | 06 September 2017, 23:48:41 UTC |
46cd357 | Project Philly | 06 September 2017, 19:07:30 UTC | Integrate liqun/CSEvalQuickNugetVersionFix into master | 06 September 2017, 19:07:30 UTC |
41691c2 | Project Philly | 06 September 2017, 18:43:12 UTC | Integrate sptiwari/fasterrcnn_cython_py35_36_binaries into master | 06 September 2017, 18:43:12 UTC |
6c6015e | Ivan Rodriguez | 28 August 2017, 16:04:14 UTC | Adding examples for output intermediate layers and combine outputs in cpp and cs | 06 September 2017, 14:26:49 UTC |
1e8b392 | Ivan Rodriguez | 28 August 2017, 16:04:14 UTC | Adding name to features and last avg pooling nodes | 06 September 2017, 11:58:36 UTC |
a62e793 | duli1 | 06 September 2017, 04:34:18 UTC | Merge branch 'duli/tensorboard_image1' of https://github.com/Microsoft/cntk into duli/tensorboard_image1 | 06 September 2017, 04:34:18 UTC |
42b2e68 | duli1 | 06 September 2017, 04:08:01 UTC | Squashed commit of the following: commit fb1ef6a11cd5d31484c0a229f2ec813ddc885a08 Author: duli1 <duli1@DU-LI-WS> Date: Fri Aug 25 10:02:23 2017 -0700 Tensorboard image feature commit a5584c245b6c8d6b4b31bacd8ac1dc3ed9242463 Author: duli1 <duli1@DU-LI-WS> Date: Fri Aug 25 10:02:23 2017 -0700 Tensorboard image feature | 06 September 2017, 04:31:31 UTC |
6319838 | duli1 | 06 September 2017, 04:08:01 UTC | Squashed commit of the following: commit fb1ef6a11cd5d31484c0a229f2ec813ddc885a08 Author: duli1 <duli1@DU-LI-WS> Date: Fri Aug 25 10:02:23 2017 -0700 Tensorboard image feature commit a5584c245b6c8d6b4b31bacd8ac1dc3ed9242463 Author: duli1 <duli1@DU-LI-WS> Date: Fri Aug 25 10:02:23 2017 -0700 Tensorboard image feature | 06 September 2017, 04:08:01 UTC |
8fc736b | Philipp Kranen | 05 September 2017, 23:06:50 UTC | Updated Examples section wrt object detection news | 05 September 2017, 23:06:50 UTC |
03bfe31 | liqfu | 05 September 2017, 21:33:28 UTC | Revert nuget version changes to Eval examples. It should be done to the release branch first and merge to master after release. | 05 September 2017, 21:33:28 UTC |
18a9f21 | Spandan Tiwari | 05 September 2017, 19:25:49 UTC | Adding Python 3.5 and 3.6 cython modules for Linux needed for FastRCNN and FasterRCNN. | 05 September 2017, 19:25:49 UTC |
7261048 | Cheng Tang | 01 September 2017, 20:50:30 UTC | fix comments in CR | 05 September 2017, 16:26:43 UTC |
b045112 | Cheng Tang | 31 August 2017, 22:52:10 UTC | reuse shape code;fix a typo | 05 September 2017, 16:26:42 UTC |
7fe0993 | Cheng Tang | 30 August 2017, 00:24:32 UTC | use reshape instead of resize | 05 September 2017, 16:26:42 UTC |
60e394b | Cheng Tang | 25 August 2017, 21:48:23 UTC | fix to_batch grad issue | 05 September 2017, 16:26:42 UTC |
23a7846 | Project Philly | 04 September 2017, 21:30:44 UTC | Integrate t-ivrodr/using_filewrapper_in_bin_reader into master | 04 September 2017, 21:30:44 UTC |
6f98774 | Ivan Rodriguez | 09 August 2017, 13:01:00 UTC | Addressing comments in CR | 04 September 2017, 12:26:46 UTC |
8e078eb | Ivan Rodriguez | 08 August 2017, 08:25:47 UTC | Use FileWrapper for file handling | 04 September 2017, 12:26:46 UTC |
22762e3 | Nikola Milosavljevic | 11 July 2017, 14:23:14 UTC | Expose crop layer in Python API | 04 September 2017, 07:19:12 UTC |
ea788c8 | liqun fu | 03 September 2017, 04:30:27 UTC | Add C# training API second commit plus Mark Hillebrand's fix to TestDriver.py | 03 September 2017, 04:30:27 UTC |
49cbd85 | Wolfgang Manousek | 31 August 2017, 11:31:51 UTC | remove cub download hash | 31 August 2017, 11:31:51 UTC |
c53bd26 | Emad Barsoum | 31 August 2017, 05:10:13 UTC | Merge branch 'master' into ebarsoum/deterministic_pool Merge with master. | 31 August 2017, 05:10:13 UTC |
bf131dd | Emad Barsoum | 31 August 2017, 00:04:56 UTC | Add deterministic flag to max pooling only. | 31 August 2017, 00:04:56 UTC |
c29b53d | Emad Barsoum | 30 August 2017, 18:12:06 UTC | pass deterministoc flag to pool node. | 30 August 2017, 18:12:06 UTC |
37854b4 | Project Philly | 30 August 2017, 18:03:56 UTC | Integrate pkranen/odScripts into master | 30 August 2017, 18:03:56 UTC |
f78adb7 | Emad Barsoum | 30 August 2017, 16:51:43 UTC | Add support to deterministic pooling. | 30 August 2017, 16:51:43 UTC |
46452cf | Philipp Kranen | 30 August 2017, 15:16:29 UTC | added Faster R-CNN to samples.json | 30 August 2017, 15:16:29 UTC |
6348856 | Philipp Kranen | 22 August 2017, 07:46:27 UTC | enabling native proposal layer and dlib selective search | 30 August 2017, 15:12:35 UTC |
7329e94 | Philipp Kranen | 02 August 2017, 15:07:35 UTC | streamlining object detection | 30 August 2017, 15:12:33 UTC |
ddcf498 | Zhou Wang | 30 August 2017, 14:13:40 UTC | update baseline file | 30 August 2017, 14:13:40 UTC |
04f52ab | Zhou Wang | 30 August 2017, 05:32:36 UTC | use Task and bind thread to model in CSEvalExamples | 30 August 2017, 11:16:02 UTC |
63ea594 | Project Philly | 30 August 2017, 09:14:05 UTC | Integrate zhouwang/eval-cppexample into master | 30 August 2017, 09:14:05 UTC |
3a288fd | Project Philly | 30 August 2017, 08:20:22 UTC | Integrate FDecaYed/dilatedConv_cudnn_V6 into master | 30 August 2017, 08:20:22 UTC |
af30726 | KGWANGMIN | 30 August 2017, 03:36:26 UTC | Fix typos (#2275) | 30 August 2017, 03:36:26 UTC |