| refs/heads/eldak/codeGen4 | 797df11 | Some experimentation | 03 August 2017, 12:29:49 UTC |
| refs/heads/eldak/compositeInMemory | aeb1a48 | Adding keep data in memory | 26 April 2017, 14:31:47 UTC |
| refs/heads/eldak/exposingCWrapperForHalide | f0d63b2 | Exposing EvaluatorWrapper interface and restructure the code to be able to reuse most of the stuff in Halide inference engine | 30 November 2017, 12:33:49 UTC |
| refs/heads/eldak/exposingGPUThruC | 8066621 | Fixing BPTT for case when the minibatch size changes | 03 January 2018, 16:40:52 UTC |
| refs/heads/eldak/exposingSequenceIdFromMinibatchData | 2518fde | Adding ability to access sequence ids | 06 June 2017, 14:54:34 UTC |
| refs/heads/eldak/fixingBatchNorm | 632cc47 | Fixing cudnn in case of 1-D arrays | 12 May 2017, 17:43:56 UTC |
| refs/heads/eldak/fixingBlockReplace | 4b7b7a8 | Fixing restore in place with wrongly reordered placeholders | 28 November 2017, 12:55:48 UTC |
| refs/heads/eldak/fixingCrop | a374e46 | Another fix | 28 July 2017, 17:28:24 UTC |
| refs/heads/eldak/fixingMaskForEval | accac8a | Fixing Mask for eval | 07 December 2017, 11:41:33 UTC |
| refs/heads/eldak/fixingUDInitialization | f98f2f3 | Fixing UD randomizer initialization | 20 October 2017, 10:12:27 UTC |
| refs/heads/eldak/forJacob | 2a1e3e7 | Adapting the test to use v1 cntk model | 18 December 2017, 11:42:17 UTC |
| refs/heads/eldak/gpuPrefetchGpuTranspose | be8b607 | Small refactoring | 15 September 2016, 12:02:07 UTC |
| refs/heads/eldak/includingCNTKLibraryC | fd74a5c | Adding C header to the drop | 05 December 2017, 08:46:22 UTC |
| refs/heads/eldak/investigatingBM | 1d824f4 | Adding report of the learning rate per sample | 21 August 2017, 10:13:40 UTC |
| refs/heads/eldak/jdEvalC | 07fcce9 | First version of C API for evaluation, taking over Jacob's branch. | 02 November 2017, 22:09:05 UTC |
| refs/heads/eldak/maxNumSequences | 47c19c5 | Merge branch 'eldak/maxNumSequences' of https://github.com/Microsoft/CNTK into eldak/maxNumSequences | 05 December 2016, 09:28:39 UTC |
| refs/heads/eldak/muditsChanges | 9771dc4 | Updating 1bitsgd | 29 August 2017, 17:28:23 UTC |
| refs/heads/eldak/muditsChanges2 | 4c0d724 | Removing some prints | 04 October 2017, 12:25:30 UTC |
| refs/heads/eldak/muditsChanges3 | 53bbd1d | Bumping 1bit version | 05 October 2017, 08:30:05 UTC |
| refs/heads/eldak/netRefactoring | af149ee | Refactoring of general graph algorithms from ComputationNetwork | 25 September 2017, 12:52:27 UTC |
| refs/heads/eldak/noRandomizerChunking | 742bf03 | Fixing no randomize to preserve the chunk while accessing sequence data | 01 September 2016, 11:46:04 UTC |
| refs/heads/eldak/opencv32 | 3c803e2 | Fixing more | 26 April 2017, 14:43:33 UTC |
| refs/heads/eldak/optimizeO | 31aa348 | Trying o3 optimization | 20 April 2017, 13:48:08 UTC |
| refs/heads/eldak/optimizingMBSize | c13efcc | Some changes | 24 March 2017, 14:21:53 UTC |
| refs/heads/eldak/optimizingSparseRead | f96aa31 | Compilation fixes for CPU mode | 10 May 2017, 13:53:08 UTC |
| refs/heads/eldak/parallelPacking | aa98ff0 | Fixing compilation | 09 May 2017, 12:15:58 UTC |
| refs/heads/eldak/refactoringForInference | 3cb15dc | Some fixes | 11 October 2017, 09:24:53 UTC |
| refs/heads/eldak/refactoringForInference2 | 0518eb4 | Fixing merging issues | 24 October 2017, 15:01:46 UTC |
| refs/heads/eldak/renamingManual | ea9d7b3 | Renaming manual | 03 July 2017, 16:20:21 UTC |
| refs/heads/eldak/retryToOpenFile | 4e87707 | Adding retry logic to open file | 20 March 2017, 12:37:13 UTC |
| refs/heads/eldak/sequenceKeyLength | 5446f71 | Fixing the length | 18 April 2017, 13:52:16 UTC |
| refs/heads/eldak/tempFixInVectorMax | 49ac3a8 | Removing debug info | 11 January 2017, 19:33:58 UTC |
| refs/heads/eldak/ud | 57bc5fb | More fixes | 26 June 2017, 11:47:29 UTC |
| refs/heads/eldak/udRandom | fd518f6 | Fixing multithreading | 10 July 2017, 14:16:32 UTC |
| refs/heads/eldak/udRandom2 | ea5a77e | Exposing user deserializer in Python | 08 June 2017, 13:39:36 UTC |
| refs/heads/fmegen/eval-internal-test | cda7fb0 | more output | 17 November 2016, 09:12:38 UTC |
| refs/heads/fmegen/eval-vs2015 | 47c3a49 | getting rid of warnings (by fixing them ;)) | 25 November 2016, 09:47:04 UTC |
| refs/heads/fmegen/eval-vs2015_2 | 7890724 | support debug libs (adding "d" at the end of a debug build and referencing debug libs when building a debug build) | 15 December 2016, 14:57:38 UTC |
| refs/heads/fmegen/linux-debian | 479b6e5 | add tests on ARM64 for ResNet_CIFAR10 move test to ARM64Test remove unneeded file add tests fix tests restore the tests here, as the tests for ARM64 are in ..\..\ARM64Tests" rename file | 27 February 2017, 15:24:10 UTC |
| refs/heads/fmegen/vs2015/changing-project-files | f353051 | Tests/EndToEndTests/Examples/Text/PennTreebank/RNN: update Windows baselines | 20 December 2016, 12:14:38 UTC |
| refs/heads/fseide/99 | deddd25 | fixed the Communicator use, which is not germaine to this tutorial (although teh bug is really in Communicator itself, which should be a dummy in non-1Bit-SGD builds) | 30 May 2017, 21:56:43 UTC |
| refs/heads/fseide/complexity | fb8c7fc | changed PrimitiveAsComposite() to directly calling CompositeFunction::Create() | 03 June 2017, 17:56:37 UTC |
| refs/heads/fseide/cudnn5 | 06b7626 | (made VC happy) | 25 August 2016, 22:07:19 UTC |
| refs/heads/fseide/dynamite | 2003303 | some updates, Marian-related | 08 September 2020, 19:38:44 UTC |
| refs/heads/fseide/e2easr | d68dc77 | lots of additions to SLUHandsOn.py | 30 September 2016, 20:26:46 UTC |
| refs/heads/fseide/fixAsMatrix | 650bad1 | bug fix: AsTensorViewShape() should not pad shapes to >= 2 axes, instead AsMatrix() should be updated to handle rank-1 tensors; bug fix: RandomSampleInclusionFrequencyNode should not return a matrix with 1 column but rather a vector proper | 04 April 2017, 20:48:19 UTC |
| refs/heads/fseide/fixLoad | c48e1a4 | merged with pkranen/doc | 29 January 2016, 20:27:10 UTC |
| refs/heads/fseide/fixSamples | 8a0974e | fixed an incorrect consistency check in TrainOrAdaptModel() | 29 January 2016, 18:53:16 UTC |
| refs/heads/fseide/kdd | 9d6345b | Merge branch 'master' of https://github.com/Microsoft/CNTK into fseide/kdd | 24 August 2016, 15:05:13 UTC |
| refs/heads/fseide/ndops | c7541b2 | changed m_consumers | 15 May 2017, 06:14:44 UTC |
| refs/heads/fseide/netlib | 89e0958 | License change | 18 January 2016, 08:33:34 UTC |
| refs/heads/fseide/pathPrefix | d923df5 | License change | 18 January 2016, 08:33:38 UTC |
| refs/heads/fseide/refactor_nodes | 7ae9d0e | License change | 18 January 2016, 08:33:42 UTC |
| refs/heads/fseide/s2s | cb61f42 | Merge branch 'master' of https://github.com/Microsoft/CNTK into fseide/s2s | 03 March 2016, 17:36:14 UTC |
| refs/heads/fseide/s2sd | c98e0ed | FlattenDims2 | 23 February 2016, 21:20:20 UTC |
| refs/heads/fseide/s2sd5 | 7f7eb8b | (comment) | 04 March 2016, 04:22:48 UTC |
| refs/heads/fseide/s2spy | 6341446 | Merge branch 'master' of https://github.com/Microsoft/cntk into fseide/s2spy | 13 April 2017, 22:23:11 UTC |
| refs/heads/fseide/seqtest | a3e2a08 | License change | 18 January 2016, 08:33:51 UTC |
| refs/heads/fseide/testlstm | 21b7649 | fixed naming of LSTM test | 02 May 2016, 23:49:00 UTC |
| refs/heads/gaizna/cntk_readerless | de6c89f | This commit changes which field to look for the "evalreader" specialized reader in the cntk config file. | 21 April 2016, 09:51:07 UTC |
| refs/heads/gaizna/eval_import_lib | 13a2b4b | Replaces loading EvalDLL dynamically, with import library. | 11 July 2016, 13:07:01 UTC |
| refs/heads/gaizna/writeminibatch_fix | 27f2858 | This commit fixes the WriteMiniBatchWithFormatting method to correctly write the matrix. Prior, it was not writing all the columns. | 25 April 2016, 13:47:10 UTC |
| refs/heads/gather_from_target | caf56bb | update UT baseline | 30 March 2017, 17:28:40 UTC |
| refs/heads/gather_op | 942a6a4 | gather gpu impl and ut | 15 March 2017, 20:30:54 UTC |
| refs/heads/giwye/lightrnn | 2882f1c | added test | 27 June 2017, 00:55:33 UTC |
| refs/heads/gloch/fix_native_proposallayerlib | d1e2c91 | Fix native ProposalLayerLib dimensions indexes | 24 April 2018, 14:16:20 UTC |
| refs/heads/gorand/lstm_experiments | ab6f3f9 | Linear tanh and sigmoid | 08 August 2017, 17:25:45 UTC |
| refs/heads/guoli/EMBR | f7b3d7d | embr | 22 December 2017, 00:03:13 UTC |
| refs/heads/guoye/EMBR | 80728b2 | embr | 20 December 2017, 01:58:52 UTC |
| refs/heads/guoye/EMBR1 | 571e787 | embr | 16 May 2018, 22:50:44 UTC |
| refs/heads/guoye/RNNT_eval_cuda10_embr_1 | 1fc15ed | multithread GPU code checkin, no speedup though due to the stream in kernel function | 21 February 2020, 06:34:44 UTC |
| refs/heads/guoye/RNNT_eval_cuda10_embr_hao_decoder | 9c720b0 | just mask one line of debug info | 26 April 2020, 15:57:58 UTC |
| refs/heads/guoye/dcfsmn | b2dee0a | fix memory issue | 04 October 2018, 03:05:09 UTC |
| refs/heads/guoye/dcfsmn1 | 12c67f6 | support semi-orthogonal | 19 October 2018, 05:00:30 UTC |
| refs/heads/guoye/dcfsmn1_debug | 35cfbb1 | debug | 22 October 2018, 22:26:43 UTC |
| refs/heads/guoye/dcfsmn2 | 3a1457b | semi-orthogonal | 26 October 2018, 00:02:35 UTC |
| refs/heads/guoye/dcfsmn4 | 7beefcb | bug fix | 19 December 2018, 07:44:30 UTC |
| refs/heads/guoye/dcfsmn5 | c10fb0c | cherry-pick 8deb528b7298f988a614f8570ef1211ccdc446d6 from t-kax/debug-fsmn cherry-pick 8deb528b7298f988a614f8570ef1211ccdc446d6 from t-kax/debug-fsmn cherry-pick 8deb528b7298f988a614f8570ef1211ccdc446d6 from t-kax/debug-fsmn fix compilation error fix compilation error 2 fix compilation error 3 fix compilation error 4 fix compilation error 5 FSMN bug fixed!! I have wrong understanding of one Minibatch Matrix's layout before and Now I have fixed related problem try to fix memory bug(illegal memory access) use MatrixPool and UpdateFunctionMBSize() to Resize m_flags & construct m_flags in BeginForwardProp() [BiVfsmnNode] avoid CPU GPU transfer but still use hard code minibatch size semi-orthogonal semi-orthogonal nan-debug nan-debug nan-debug nan-debug nan-debug semi-orthogonal-errinfo semi-orthogonal alpha-debug semi-orthogonal-errinfo semi-orthogonal semi-orthogonal-errinfo semi-orthogonal semi-orthogonal-errinfo nan-debug nan-debug nan-debug nan-debug nan-debug nan-debug-improve nan-fix update bo's fix, use seqIdx based FSMN flags bug fix bug fix mask unreferenced local virable LR in SGD.cpp | 11 June 2018, 10:05:42 UTC |
| refs/heads/guoye/dcfsmn6 | bb8027f | continue to change code to make LR=0 | 02 April 2019, 18:54:51 UTC |
| refs/heads/guoye/master_debug | dcd4aef | master_debug | 31 December 2017, 06:02:11 UTC |
| refs/heads/guoye/merge_embr | 5771cb8 | embr | 07 September 2018, 19:59:33 UTC |
| refs/heads/guoye/merge_embr_final | b568283 | make the function interface of AssignSequenceError in NoGpu.cpp to be consistent with GPUMatrix.h | 22 January 2019, 23:16:42 UTC |
| refs/heads/guoye/merge_embr_final_v1 | 777c3ae | remove space | 11 February 2019, 19:43:35 UTC |
| refs/heads/hecli/PostReleaseMerge | c3e4f66 | 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 |
| refs/heads/hjl/Char-RNN-Example | 63b0878 | add script to run Char-RNN example | 04 February 2016, 01:31:42 UTC |
| refs/heads/huhu/RNNT_SR_master | 1fd8785 | RNNT init7 | 11 July 2019, 01:18:50 UTC |
| refs/heads/hunter/fast-rcnn | 29dffa7 | add brainscript version, update configs. | 19 August 2016, 20:52:13 UTC |
| refs/heads/hunter/fast-rcnn2 | fc50754 | address review comments, round 1 | 12 August 2016, 22:27:07 UTC |
| refs/heads/hunter/fix-write | 71b012e | Fix write action to correctly modify Reader config | 17 June 2016, 13:59:54 UTC |
| refs/heads/issaca | a737b47 | Merge branch 'master' into issaca | 02 June 2016, 19:53:17 UTC |
| refs/heads/iu.triplet_loss | c8df6e8 | Merged with master | 08 December 2016, 02:01:30 UTC |
| refs/heads/iu_layer | 6cb945e | Remove debug info for triplet | 21 October 2016, 14:38:14 UTC |
| refs/heads/ivanst/profiler | da06711 | Use std::mutex | 29 December 2016, 15:38:42 UTC |
| refs/heads/ivanst/profiler2 | 9603e5f | Fix Linux build | 26 May 2016, 13:17:48 UTC |
| refs/heads/ivanst/profiler3 | 2e2b5fc | Add more detailed perf events for gradient aggregation | 30 May 2016, 18:10:10 UTC |
| refs/heads/ivanst/profiler5 | 7f0ab8d | Add additional markers for ImageReader' | 10 June 2016, 23:32:39 UTC |
| refs/heads/ivanst/profiler_20160510 | 11b96f8 | More profiling buffer fixes and and speed improvements. | 10 May 2016, 21:29:56 UTC |
| refs/heads/ivanst/profiler_20161108 | ad1be4c | CNTK performance profiler Equivalent to zsoltm/profiler as of 2016-11-08. | 08 November 2016, 22:15:55 UTC |
| refs/heads/ivanst/profiler_chr | 14d37fc | QP vs. chrono | 27 December 2016, 16:50:22 UTC |
| refs/heads/ivanst/profiler_cr | ce767fe | Address CR feedback (Iteration 9) | 23 December 2016, 17:51:08 UTC |
