| f292911 | William Darling | 17 February 2017, 13:23:05 UTC | updated to be compatible w beta 11 | 17 February 2017, 13:23:05 UTC |
| 9931b3a | William Darling | 17 February 2017, 13:22:47 UTC | fixed some issues | 17 February 2017, 13:22:47 UTC |
| 8f18ffe | William Darling | 24 January 2017, 14:49:27 UTC | also save a base version of the model | 24 January 2017, 14:49:27 UTC |
| b6defc5 | William Darling | 24 January 2017, 14:45:28 UTC | included python-based (probably very slow) beam search implementation | 24 January 2017, 14:45:28 UTC |
| da2dd66 | William Darling | 23 January 2017, 12:57:23 UTC | small change to validation mb data due to how the reader works now | 23 January 2017, 12:57:23 UTC |
| e362817 | William Darling | 19 January 2017, 10:44:41 UTC | some fixes to make sure it works when not using an embedding and when not using attention | 19 January 2017, 10:44:41 UTC |
| 33e4683 | William Darling | 17 January 2017, 10:51:29 UTC | work-around for find_by_name bug | 17 January 2017, 10:51:29 UTC |
| 56d35c3 | William Darling | 17 January 2017, 10:31:52 UTC | Merge branch 'wdarling/lstmaux' of https://github.com/Microsoft/CNTK into wdarling/lstmaux Conflicts: Examples/SequenceToSequence/CMUDict/Python/Sequence2Sequence.py | 17 January 2017, 10:31:52 UTC |
| e8a412f | William Darling | 17 January 2017, 10:23:22 UTC | some updates to keep up with the changes | 17 January 2017, 10:23:22 UTC |
| 8bc9ac6 | William Darling | 14 December 2016, 10:50:50 UTC | moved back to non-local blocks | 14 December 2016, 10:50:50 UTC |
| 05dd211 | William Darling | 13 December 2016, 14:43:29 UTC | added embedding of inputs | 13 December 2016, 14:43:29 UTC |
| 7878eed | William Darling | 06 December 2016, 12:58:12 UTC | back to correct blocks lib | 06 December 2016, 12:58:12 UTC |
| 87729e3 | William Darling | 06 December 2016, 12:56:53 UTC | updated test to be more precise; made some changes to test function to work more nicely | 06 December 2016, 12:56:53 UTC |
| cc1bbc0 | William Darling | 06 December 2016, 12:31:34 UTC | adding comments to clarify the role of the decoder and where it is part of the model vs where it's not | 06 December 2016, 12:31:34 UTC |
| 83f61e3 | William Darling | 06 December 2016, 12:09:22 UTC | take model loading out of write function | 06 December 2016, 12:09:22 UTC |
| cb39afc | William Darling | 06 December 2016, 11:04:31 UTC | added TODO for adding axis to softmax | 06 December 2016, 11:04:31 UTC |
| 67f1514 | William Darling | 30 November 2016, 10:28:50 UTC | moved input modeling outside of the create_model() function | 30 November 2016, 10:28:50 UTC |
| 8a74cc1 | William Darling | 30 November 2016, 09:42:46 UTC | Merge branch 'wdarling/lstmaux' of https://github.com/Microsoft/cntk into wdarling/lstmaux Conflicts: Examples/SequenceToSequence/CMUDict/Python/Sequence2Sequence.py Examples/SequenceToSequence/CMUDict/Python/attention.py | 30 November 2016, 09:42:46 UTC |
| 729355c | William Darling | 30 November 2016, 09:40:26 UTC | fixed attention implementation; attention is used for each layer but it of course needs to be one shared attention weight vector | 30 November 2016, 09:40:26 UTC |
| c9078bc | William Darling | 29 November 2016, 11:51:34 UTC | change behavior of zeroing-out non-used attention weights | 29 November 2016, 11:51:34 UTC |
| e9b7ebf | William Darling | 27 November 2016, 14:08:35 UTC | fixed attention visualization for interactive sessions | 27 November 2016, 14:08:35 UTC |
| 2de707f | William Darling | 27 November 2016, 13:32:15 UTC | bug fixes; changed way that attention weights are calculated | 27 November 2016, 13:32:15 UTC |
| fcf22ae | William Darling | 27 November 2016, 10:47:04 UTC | added interactive session with attention logging | 27 November 2016, 10:47:04 UTC |
| 78064ba | William Darling | 24 November 2016, 14:56:02 UTC | more CR-based changes | 24 November 2016, 14:56:02 UTC |
| d871301 | William Darling | 24 November 2016, 14:23:35 UTC | addressing more CR comments including adding an example word translation function using numpy inputs | 24 November 2016, 14:23:35 UTC |
| 9caa462 | William Darling | 23 November 2016, 16:36:24 UTC | addressed a whole bunch of CR comments | 23 November 2016, 16:36:24 UTC |
| d368435 | William Darling | 23 November 2016, 10:27:28 UTC | first set of fixes for addressing CR comments | 23 November 2016, 10:27:28 UTC |
| 8ee8f09 | William Darling | 22 November 2016, 11:30:53 UTC | point to real blocks lib | 22 November 2016, 11:30:53 UTC |
| 2e7120a | William Darling | 22 November 2016, 11:26:36 UTC | updated test for new version of seq2seq | 22 November 2016, 11:26:36 UTC |
| 4ee0027 | William Darling | 22 November 2016, 10:42:48 UTC | Merge branch 'wdarling/lstmaux' of https://github.com/Microsoft/CNTK into wdarling/lstmaux | 22 November 2016, 10:42:48 UTC |
| bbfd42c | William Darling | 22 November 2016, 10:42:03 UTC | fix paths | 22 November 2016, 10:42:03 UTC |
| db68246 | William Darling | 22 November 2016, 10:39:24 UTC | renamed Attention.py to attention.py | 22 November 2016, 10:39:24 UTC |
| 8f181f7 | William Darling | 22 November 2016, 10:25:00 UTC | added attention implementation | 22 November 2016, 10:25:00 UTC |
| 1e8309f | William Darling | 22 November 2016, 10:24:45 UTC | updated Sequence2Sequence example to use blocks/layers library and added attention | 22 November 2016, 10:24:45 UTC |
| b5c03e4 | William Darling | 22 November 2016, 10:06:51 UTC | updated blocks to support aux in LSTM | 22 November 2016, 10:06:51 UTC |
| 1861a2a | Chris Basoglu | 22 November 2016, 01:38:19 UTC | Update README.md | 22 November 2016, 01:38:19 UTC |
| dbfb1b5 | Project Philly | 21 November 2016, 22:40:42 UTC | Integrate zhouwang/evalwrapper-multioutputs into master | 21 November 2016, 22:40:42 UTC |
| f4a16b4 | Mark Hillebrand | 21 November 2016, 21:08:20 UTC | Merge remote-tracking branch 'origin/master' into mahilleb/PostReleaseMerges | 21 November 2016, 21:08:20 UTC |
| b22b4bd | Mark Hillebrand | 21 November 2016, 21:00:12 UTC | Revert "add hyper parameter tuning tool as a sub module to CNTK." This reverts commit 4442bfc1d5688d565b526d0b4026fc84182155ef. | 21 November 2016, 21:00:12 UTC |
| 4442bfc | Dong Yu | 21 November 2016, 20:03:46 UTC | add hyper parameter tuning tool as a sub module to CNTK. | 21 November 2016, 20:03:46 UTC |
| cc9a5e7 | Zhou Wang | 17 November 2016, 18:34:12 UTC | added unit tests and update baseline | 21 November 2016, 19:23:44 UTC |
| 971778d | Zhou Wang | 17 November 2016, 13:17:06 UTC | fix wrong mapping between node name and value in case of multiple outputs | 21 November 2016, 19:20:23 UTC |
| 75c4229 | Chris Basoglu | 21 November 2016, 18:53:52 UTC | Update tutorials | 21 November 2016, 18:53:52 UTC |
| 7d3ec64 | Zhou Wang | 21 November 2016, 13:53:20 UTC | update baseline | 21 November 2016, 14:54:52 UTC |
| f043a67 | Anton Schwaighofer | 21 November 2016, 13:16:38 UTC | Removed testing for Outcome | 21 November 2016, 14:54:52 UTC |
| 7fbb791 | Anton Schwaighofer | 21 November 2016, 12:54:27 UTC | corrected yml test setup | 21 November 2016, 14:54:52 UTC |
| ef18e10 | Anton Schwaighofer | 21 November 2016, 12:40:52 UTC | added output cardinality tests | 21 November 2016, 14:54:51 UTC |
| 1faa42f | Anton Schwaighofer | 21 November 2016, 10:30:23 UTC | Moved exception testing to test suite. Addressed all but 1 code review comment. Perf change also for ParallelExtractHwc. | 21 November 2016, 14:54:51 UTC |
| 9573060 | Anton Schwaighofer | 18 November 2016, 13:28:18 UTC | Adding test baseline results | 21 November 2016, 14:54:51 UTC |
| 46a56fc | Anton Schwaighofer | 18 November 2016, 13:18:00 UTC | removed hard tabs | 21 November 2016, 14:54:51 UTC |
| dfbe059 | Anton Schwaighofer | 18 November 2016, 11:40:55 UTC | updated references | 21 November 2016, 14:54:51 UTC |
| 45ceb56 | Anton Schwaighofer | 18 November 2016, 11:37:34 UTC | Cleaned up image API call, turned into test suite | 21 November 2016, 14:54:51 UTC |
| 97e60b5 | Anton Schwaighofer | 17 November 2016, 14:17:48 UTC | updated according to CR comments | 21 November 2016, 14:54:51 UTC |
| 896837f | Anton Schwaighofer | 11 November 2016, 13:10:32 UTC | Changes to EvalWrapper, should not be in final checkin | 21 November 2016, 14:54:50 UTC |
| d39a2b6 | Anton Schwaighofer | 26 October 2016, 10:19:30 UTC | removing surplus namespace opening | 21 November 2016, 14:54:50 UTC |
| 98107c7 | Anton Schwaighofer | 26 October 2016, 10:03:41 UTC | Changing loop structure to speed up | 21 November 2016, 14:54:50 UTC |
| 86863f6 | Anton Schwaighofer | 26 October 2016, 10:03:11 UTC | added changes requested in code review | 21 November 2016, 14:54:50 UTC |
| a0b0a5c | Anton Schwaighofer | 25 October 2016, 14:33:29 UTC | Cleaned up code, remove debug outputs | 21 November 2016, 14:54:50 UTC |
| ecc13f9 | Anton Schwaighofer | 25 October 2016, 13:35:44 UTC | first version of new EvaluateImage API, still with lots of debug output | 21 November 2016, 14:54:49 UTC |
| 09f2dfd | Mark Hillebrand | 21 November 2016, 13:58:15 UTC | Examples/Image/Classification/ResNet/Python/README.md: tune | 21 November 2016, 14:22:28 UTC |
| 8c27d65 | Mark Hillebrand | 21 November 2016, 13:56:09 UTC | Grammar | 21 November 2016, 14:22:20 UTC |
| d3f3235 | Mark Hillebrand | 21 November 2016, 13:54:25 UTC | Examples/Image/Classification/ResNet/Python/run_Distributed.py: allow to run on Linux | 21 November 2016, 14:22:12 UTC |
| f0ebffc | Mark Hillebrand | 21 November 2016, 12:57:33 UTC | Examples/Image/Classification/*/Python/: fix broken example on Linux (wrong path data sets) | 21 November 2016, 14:22:03 UTC |
| afd0a87 | Mark Hillebrand | 21 November 2016, 09:46:21 UTC | Scripts/install/linux/install-cntk.sh: fix conda env update | 21 November 2016, 14:21:43 UTC |
| bbc8c6a | Mark Hillebrand | 19 November 2016, 14:23:53 UTC | bindings/python/doc/gettingstarted.rst: fix example | 21 November 2016, 14:21:43 UTC |
| a80deb4 | Mark Hillebrand | 19 November 2016, 13:54:14 UTC | bindings/python/doc/examples.rst: updated link for CIFAR ResNet | 21 November 2016, 14:21:43 UTC |
| 582420e | Mark Hillebrand | 18 November 2016, 22:14:58 UTC | Scripts/install/linux/install-cntk.sh: make executable | 21 November 2016, 14:21:43 UTC |
| ce301e2 | Mark Hillebrand | 18 November 2016, 18:44:02 UTC | Bump version | 21 November 2016, 14:21:43 UTC |
| 746eb2d | Eldar Akchurin | 21 November 2016, 11:26:57 UTC | Fixing baseline | 21 November 2016, 11:26:57 UTC |
| 6572d53 | Eldar Akchurin | 18 November 2016, 16:42:08 UTC | Fixing defaults for the composite reader/image deserializer | 21 November 2016, 09:14:06 UTC |
| bcf05b4 | Thilo Will | 19 November 2016, 20:53:49 UTC | Merge branch 'thilow/FixDenseTimesSparseAsBlocksparse4' of https://github.com/Microsoft/cntk | 19 November 2016, 20:53:49 UTC |
| a06c88c | Thilo Will | 19 November 2016, 20:51:42 UTC | added TempDeviceBuffer to CommonMatrix.h | 19 November 2016, 20:53:24 UTC |
| 913dcd3 | Thilo Will | 18 November 2016, 22:57:54 UTC | fixes for dense times sparse | 19 November 2016, 20:53:24 UTC |
| 72ba3d3 | Thilo Will | 19 November 2016, 20:51:42 UTC | added TempDeviceBuffer to CommonMatrix.h | 19 November 2016, 20:51:42 UTC |
| d60e397 | Mark Hillebrand | 19 November 2016, 13:38:31 UTC | Update wrt. deleted examples; cifar_convnet_distributed_test.py: skip on Windows | 19 November 2016, 13:38:31 UTC |
| d413fdf | KeDengMS | 19 November 2016, 02:51:13 UTC | Fix run_Distributed to use the right script | 19 November 2016, 02:51:13 UTC |
| 13e1626 | KeDengMS | 19 November 2016, 02:29:58 UTC | Merge branch 'master' into kedeng/distExample | 19 November 2016, 02:29:58 UTC |
| aebe931 | KeDengMS | 19 November 2016, 00:42:52 UTC | Fix broken ConvNet examples | 19 November 2016, 00:42:52 UTC |
| 8dec190 | Project Philly | 19 November 2016, 00:40:19 UTC | Integrate ebarsoum/convpad into master | 19 November 2016, 00:40:19 UTC |
| 95fa261 | KeDengMS | 19 November 2016, 00:03:18 UTC | Update test expectation after reduced epochs | 19 November 2016, 00:03:18 UTC |
| 07db03a | KeDengMS | 18 November 2016, 23:26:21 UTC | Update README.md | 18 November 2016, 23:26:21 UTC |
| 19b357c | KeDengMS | 18 November 2016, 23:19:25 UTC | Tweak text and test | 18 November 2016, 23:19:25 UTC |
| ea432d7 | KeDengMS | 18 November 2016, 23:12:25 UTC | Update ConvNet example and test | 18 November 2016, 23:12:25 UTC |
| 085ecbe | Emad Barsoum | 18 November 2016, 23:00:43 UTC | replace all global default with its accessor | 18 November 2016, 23:00:43 UTC |
| 160248e | Thilo Will | 18 November 2016, 22:57:54 UTC | fixes for dense times sparse | 18 November 2016, 22:57:54 UTC |
| 7657d9d | Emad Barsoum | 18 November 2016, 22:54:54 UTC | Fix default option. | 18 November 2016, 22:54:54 UTC |
| a87795b | KeDengMS | 18 November 2016, 22:23:24 UTC | Update resnet python samples with distributed | 18 November 2016, 22:23:24 UTC |
| 7e325c8 | KeDengMS | 18 November 2016, 19:13:38 UTC | Fix Multiverso build error in Release_NoOpt | 18 November 2016, 19:13:38 UTC |
| b052cec | Mark Hillebrand | 18 November 2016, 18:15:15 UTC | Tutorials: more fixes | 18 November 2016, 18:15:15 UTC |
| 999a38a | Project Philly | 18 November 2016, 18:02:47 UTC | Integrate mahilleb/PySamplesTune into master | 18 November 2016, 18:02:47 UTC |
| 4882628 | Eldar Akchurin | 18 November 2016, 15:35:29 UTC | Merging master | 18 November 2016, 15:35:29 UTC |
| 16da600 | Project Philly | 18 November 2016, 14:50:26 UTC | Integrate eldak/increaseStackSize into master | 18 November 2016, 14:50:26 UTC |
| e442dd4 | Mark Hillebrand | 18 November 2016, 10:48:30 UTC | Tutorials: more fixes | 18 November 2016, 13:47:17 UTC |
| 6d720f3 | Eldar Akchurin | 18 November 2016, 13:16:23 UTC | Removing block momentum example for now | 18 November 2016, 13:16:23 UTC |
| efb155b | Eldar Akchurin | 18 November 2016, 11:21:25 UTC | Increasing stack size for tests | 18 November 2016, 12:23:12 UTC |
| f5f2224 | Willi Richert | 18 November 2016, 11:16:59 UTC | Fix NDMask shapes; allow sparse data with dense inputs; skip sparse tests on GPU | 18 November 2016, 11:21:18 UTC |
| 49125b5 | Willi Richert | 17 November 2016, 19:12:44 UTC | Adapting to stricter shape checks | 18 November 2016, 11:21:17 UTC |
| 41bf56c | Willi Richert | 17 November 2016, 16:51:56 UTC | Fix tests after rebase | 18 November 2016, 11:21:16 UTC |
| aade2b5 | Willi Richert | 17 November 2016, 15:24:06 UTC | Clean up | 18 November 2016, 11:21:14 UTC |
| 9c2ecb7 | Willi Richert | 17 November 2016, 09:34:54 UTC | Make training_parameter_schedule more robust | 18 November 2016, 11:21:13 UTC |
