HEAD | e939648 | disable fp16 test for poolWithSequenceAxis (#3810) | 31 March 2020, 15:55:14 UTC |
refs/heads/2.0rc/breakingChanges | 0ee09cf | Merge branch 'master' into 2.0rc/breakingChanges For checkin | 28 March 2017, 01:11:12 UTC |
refs/heads/Aghagolzadeh-brainscript_useV2Aggregator_true | d85164c | brainscript_useV2Aggregator_true static bool UseV2Aggregator() { return true; } | 16 October 2018, 17:42:29 UTC |
refs/heads/Aghagolzadeh-dockerfile_AzurebatchAI | e82e9f7 | dockerfile_AzurebatchAI docker file copied from https://github.com/Azure/BatchAI/blob/master/recipes/CNTK/CNTK-GPU-Python-Distributed-Infiniband/dockerfile instead of https://github.com/Microsoft/CNTK/edit/master/Tools/docker/CNTK-GPU-Image/Dockerfile | 16 October 2018, 21:58:26 UTC |
refs/heads/BMUF-float16 | c298985 | Update progress_print.py will add absolute time and date to each log output | 19 November 2018, 20:16:07 UTC |
refs/heads/BMUF_fp16 | 5905b2f | Update BlockMomentumDistributedLearner.h updated reset function for BMUF to support half precision | 25 April 2018, 18:43:13 UTC |
refs/heads/CUDA-elementwise-rework | 6e118f3 | License change | 18 January 2016, 08:31:35 UTC |
refs/heads/Compliance2.2 | 15d751c | fix rest issues find with code scan | 29 September 2017, 02:01:29 UTC |
refs/heads/Dev | a4760fd | License change | 18 January 2016, 08:31:39 UTC |
refs/heads/KaldiReaderOnWindows | f460b5f | License change | 18 January 2016, 08:31:42 UTC |
refs/heads/LastKnownGood | 59d382f | License change | 18 January 2016, 08:31:47 UTC |
refs/heads/MasterBeforeRemovingUnusedCode | 1aefd22 | Change default targets for build-and-test | 22 January 2016, 14:46:46 UTC |
refs/heads/aayushg/Seq2SeqRepro | 1e1a589 | Repro changes | 25 November 2016, 16:06:45 UTC |
refs/heads/aayushg/autoencoder | f0dab7b | Changes | 10 November 2016, 14:59:56 UTC |
refs/heads/aayushg/seq2seqChanges | e542ed2 | Randomization window | 25 November 2016, 15:47:18 UTC |
refs/heads/aforesti/compnetwork-constructors | abfb8b1 | Deleted unsupported constructor and = operator Replaced placeholder functions in ComputationNetwork.h with =delete's. As part of that, also removed an obsolete pragma directive. | 23 September 2016, 21:33:45 UTC |
refs/heads/alahiri/distrib-reading-LMSequenceReader | b7053a3 | reverting back to origin/master copy of CNTK.sln | 31 May 2016, 21:01:25 UTC |
refs/heads/alexeyk/bn | 357f4a8 | Fixed BN selection to handle cuDNN BN on CPU. | 01 April 2016, 23:32:31 UTC |
refs/heads/alexeyk/bn_fix | dbc5783 | Updated testcase. | 10 May 2016, 17:58:13 UTC |
refs/heads/alexeyk/conv_fix | 27f2af8 | License change | 18 January 2016, 08:31:54 UTC |
refs/heads/alexeyk/conv_tests | f6f15b6 | License change | 18 January 2016, 08:32:02 UTC |
refs/heads/alexeyk/cudnn3 | b8e1fc8 | License change | 18 January 2016, 08:32:07 UTC |
refs/heads/alexeyk/legacy | 74eca8e | Enabled legacy layout for ND convos. | 02 June 2016, 21:48:19 UTC |
refs/heads/alexeyk/logsum | 617964b | Added ReduceElementsNode to NDL. | 16 May 2016, 23:55:43 UTC |
refs/heads/alexeyk/perf | 9964783 | Changes to test perf. | 28 March 2016, 18:14:08 UTC |
refs/heads/alexeyk/tests | 1b2884d | Disabled all but reference convo engines. | 23 May 2016, 18:26:37 UTC |
refs/heads/alexeyo/1bit-SGD-License-Update-Do-Not-Integrate | b2e803f | 1bit-SGD License Update | 10 July 2016, 11:00:21 UTC |
refs/heads/alexeyo/Docker-Image | bb7638d | Next Run-Time iteration | 14 October 2016, 10:35:46 UTC |
refs/heads/alexeyo/ReadMe-News-February-10 | 54d1315 | Main ReadMe News, February 10, 2017 | 10 February 2017, 16:34:50 UTC |
refs/heads/alexeyo/fix-linux-configure-script | 5f45614 | Update GDK path in configure script. Fixes #2178 | 13 August 2017, 17:36:25 UTC |
refs/heads/alexeyo/fix-linux-install-mpi-path | 490c1e4 | Fixing OpenMPI PATH for Ubuntu 14, client script installation, including Docker Hub Images. Fixes #2191 Fixing MPI Path for Ubuntu 14 - addressing CR comments | 11 August 2017, 16:59:37 UTC |
refs/heads/alexeyo/test-winodows-drop-new-file-names | 9b0a342 | Tweaking wheel version to test drops | 27 March 2017, 13:10:01 UTC |
refs/heads/alrezni/MatrixQuantizerTestsFailRepro | c956780 | License change | 18 January 2016, 08:32:11 UTC |
refs/heads/alrezni/RandomGeneratorCpu | ac780ef | License change | 18 January 2016, 08:32:16 UTC |
refs/heads/alrezni/RandomUniform | eb89c00 | License change | 18 January 2016, 08:32:22 UTC |
refs/heads/alrezni/ReseedCudaGenerator | ddb6f9a | License change | 18 January 2016, 08:32:28 UTC |
refs/heads/alrezni/bptt_ctf | 03d01fa | Update lr_bs.cntk | 14 December 2016, 08:40:56 UTC |
refs/heads/alrezni/checkpointing_v2 | b7ddf60 | Minor clean-up Clean-up interfaces (remove smoothed gradients getter, add a reset method). Include version info in the checkpoint. | 29 June 2016, 15:09:10 UTC |
refs/heads/alrezni/ctf_nan | 28dbe0b | Add 'NaN' support to CTF | 07 June 2017, 15:48:56 UTC |
refs/heads/alrezni/cuda_reset_test | 75f6395 | Reset cuda device | 29 October 2016, 15:40:20 UTC |
refs/heads/alrezni/divering_loss_with_checkpointing | d6f33c7 | Repro for diverging loss after checkpoining | 12 October 2016, 12:32:00 UTC |
refs/heads/alrezni/fix_pytests | 5c92503 | Force set default device in pytests | 02 December 2016, 15:53:33 UTC |
refs/heads/alrezni/index_cache | 2d32149 | Refactor and simplify MLFIndexBuilder MLF indexer now also uses BufferedFileReader. Faster reading (in 512KB chunks) when loading index from cache. | 30 June 2017, 14:29:35 UTC |
refs/heads/alrezni/learners_v2 | ecfe5ff | Remove f suffix from 0.0f in LearnerBase for DateType::Double | 13 July 2016, 09:12:17 UTC |
refs/heads/alrezni/learners_v2_serialization | 9764b51 | Address CR feedback | 18 August 2016, 12:13:03 UTC |
refs/heads/alrezni/learners_v2_uts | f67cbab | Learners v2 unit tests | 04 August 2016, 15:13:52 UTC |
refs/heads/alrezni/mlf_fix | e42310c | MLFIndexBuilder: skip empty lines | 08 August 2017, 15:02:58 UTC |
refs/heads/alrezni/mpi_lib_loading | 2362898 | Do not instantiate MPI, if not running under mpiexec | 03 April 2017, 10:18:01 UTC |
refs/heads/alrezni/proposal_layer | 65471cd | Add Proposal Layer as a native UDF | 07 August 2017, 16:34:57 UTC |
refs/heads/alrezni/reader_randomization_test | 5f0af89 | Add reader randomizaiton test Add a quick python test to help visually evaluate randomization quality | 16 June 2017, 11:18:47 UTC |
refs/heads/alrezni/simple_mnist_repro | 49856ab | Revert local adjustments | 17 October 2016, 10:10:32 UTC |
refs/heads/alrezni/text_reader_benchmark | d3db246 | Merge branch 'master' into alrezni/text_reader_benchmark Conflicts: CNTK.sln | 13 May 2016, 13:36:50 UTC |
refs/heads/alrezni/v2_communicator | ec6a0a7 | Communicator: wip 3 | 24 September 2016, 02:30:11 UTC |
refs/heads/alrezni/xpacker | ee71e2c | Add XPacker | 17 August 2016, 12:15:43 UTC |
refs/heads/alrezni/xpacker2 | 54bbfe0 | Add XPacker | 17 August 2016, 12:15:43 UTC |
refs/heads/amitaga/apiChangesForRC | 91c566f | CNTK v2 library: Fix trainer to save the model instead of the combined model, loss, metric as the model file. | 04 April 2017, 20:25:02 UTC |
refs/heads/amitaga/batchNorm | e9162e1 | Synchronize batch norm nodes' running means and variances across parallel workers at end of each epoch to ensure accurate restartability across epochs and add and end-to-end test for this scenario | 08 May 2016, 01:37:50 UTC |
refs/heads/amitaga/beta11BugFixes | e3b7821 | CNTK v2 library: Change combine to be a true no-op in the V2 Function graph and implement Alias as a no-op instead of a pass-through copy | 08 February 2017, 08:44:41 UTC |
refs/heads/amitaga/bn | 7bda219 | Add arch=compute_30 codegen for debug builds required by CNTK native BatchNorm implementation | 15 February 2016, 21:03:31 UTC |
refs/heads/amitaga/cudaErrorBugFix | a6de5f8 | Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/cudaErrorBugFix | 29 April 2016, 01:45:12 UTC |
refs/heads/amitaga/cudaMemCheckForE2ETests | 6c1efa9 | Run E2E tests under CUDA memcheck | 27 June 2016, 23:32:02 UTC |
refs/heads/amitaga/dataInteropFeatures | b292e64 | CNTK v2 library: Only allocate matrices for the actual outputs desired from the CompoisteFunction | 13 January 2017, 21:54:38 UTC |
refs/heads/amitaga/fix1BitPerf | 5cc66af | CNTK parallel training: Turned off quantization for gradient matrices where row count is too small to benefit from quantization | 13 December 2016, 06:34:50 UTC |
refs/heads/amitaga/fixSimpleAggregatorPerf | 809b30c | CNTK parallel training: Enable CUDA aware MPI use for simple gradient aggregator | 13 December 2016, 21:58:26 UTC |
refs/heads/amitaga/htkmlfreaderMemoryOptimizations | 51f6db9 | Testing something | 12 February 2016, 06:29:28 UTC |
refs/heads/amitaga/mblayoutMismatchError | 2cd35ae | CNTK v2 library: Add the ability to broadcast an input without MBLayout along a specified layout, in the ReconcileDynamicAxis node | 19 January 2017, 23:37:08 UTC |
refs/heads/amitaga/memoryDebug | 0a49868 | Replace aligned alloc with unaligned alloc to debug OOM issue | 10 February 2016, 07:46:39 UTC |
refs/heads/amitaga/noDEBUG | fe6c89f | Disable _DEBUG macro for testing | 29 June 2016, 07:08:38 UTC |
refs/heads/amitaga/sparseMemAllocFix | ac266fa | Fixed a memory allocation bug in GPUSparseMatrix | 29 February 2016, 00:56:09 UTC |
refs/heads/amitaga/v2Library | d67669c | Library | 01 April 2016, 22:16:52 UTC |
refs/heads/arturl-uwp-doc | f990e8f | Update README.md | 18 July 2017, 22:49:25 UTC |
refs/heads/arturl-uwp-sample | 40a306f | use assembly reference to fix error using nuget restore | 28 July 2017, 12:18:05 UTC |
refs/heads/arturl-uwp-sample-pick-model | f235ecb | Improve UWP sample Let the user pick the model file | 09 August 2017, 19:24:48 UTC |
refs/heads/barlaso/SampledTimes | c9a122a | Added python bindings (which don't work). Cleaned up code and documentation | 27 December 2016, 03:12:46 UTC |
refs/heads/binbzha/backstitch | 0d40959 | add backstitch | 25 July 2017, 02:25:44 UTC |
refs/heads/binbzha/fsmn | 4419c2b | add bidirectional FSMN node and make it work in NDL and add FSMN CPU forward | 04 June 2018, 03:50:57 UTC |
refs/heads/binbzha/latency-control-blstm | acb9cff | refine evaluate bug | 13 November 2017, 09:19:48 UTC |
refs/heads/binbzha/latency-control-blstm-2.5 | 98a387b | refine access future error | 15 June 2018, 08:07:02 UTC |
refs/heads/binbzha/low-frame-rate | edab7b0 | add skip and delay | 24 July 2017, 04:30:57 UTC |
refs/heads/binbzha/tf-lstm-refine | 5a0dd3e | refine tf v1 | 18 April 2018, 06:50:48 UTC |
refs/heads/binxu/LeakyReLU | 3a3cdf2 | Avoid possible failure caused by boundary query. | 02 September 2016, 05:28:53 UTC |
refs/heads/bmitra/NDRMReader | 2c6f765 | Add support for resetOffsetEveryEpoch param Add support for resetOffsetEveryEpoch param | 21 September 2016, 19:39:49 UTC |
refs/heads/bmitra/fix | 5051c9e | License change | 18 January 2016, 08:33:02 UTC |
refs/heads/boren/bfsmn_binary_mlf_2.6 | 0fdc927 | bug fix | 01 November 2018, 13:27:59 UTC |
refs/heads/boren/bfsmn_binary_mlf_on_release_2.6 | b5c523d | rebased on ComputationNodeNonLooping | 30 October 2018, 03:28:45 UTC |
refs/heads/boren/bfsmn_binary_mlf_on_release_2.7 | af267cf | update onnx_repo submodule from release_2.7 | 04 June 2019, 08:19:34 UTC |
refs/heads/boren/release_2.6_kax_fsmn | 4588b87 | update the AIP of CNTKLibraryC for richland | 25 February 2019, 10:37:37 UTC |
refs/heads/bospoort-maartenb/cntk_103a | 49695c2 | Merged changes from git issue #2022 | 11 August 2017, 01:26:43 UTC |
refs/heads/bowbao/arguments_order | ea352b5 | update documentation for arguments order change | 20 July 2018, 00:57:07 UTC |
refs/heads/bowbao/cuda10 | 392e7ef | update dockerfile | 06 December 2018, 22:54:11 UTC |
refs/heads/bowbao/freedim_perf | 40af7bb | Improve conv performance with freedimension * CPU. - Short-circuit the call to ComputeConvolveGeometryExplicit() when MKL is enabled. - Replace div/mod inside ComputeConvolveGeometryExplicit with fast_divmod. * GPU. - Instead of creating a new CuDnnConvolutionEngine for every batch, just update the geometry related info, and try to reuse the workspacememory from previous run. | 18 September 2018, 01:10:34 UTC |
refs/heads/bowbao/mkl_asym_pad_and_channel_pad | 2287f5a | change default to not pad for channel in convolution. | 25 July 2018, 20:42:31 UTC |
refs/heads/bowbao/mkl_channel_pad | 407adcb | handle out channel size > 1 when checking channel pad. | 25 July 2018, 21:46:59 UTC |
refs/heads/bowbao/nightly-fix | 5594df4 | Update binary drop dependencies | 04 January 2019, 20:53:51 UTC |
refs/heads/bowbao/onnx_conv_kerneldim | ad1f8bc | Update conv input space filled by kernel check. | 31 July 2018, 22:03:09 UTC |
refs/heads/bowbao/onnx_flatten | 0c69e6b | fix Slice for onnx_backend_test | 20 August 2018, 18:55:09 UTC |
refs/heads/bowbao/onnx_gather_prelu | 113ba02 | remove print | 24 July 2018, 20:08:52 UTC |
refs/heads/bowbao/onnx_init_typetoproto | ac44642 | fix for initializing DataTypeUtils::GetTypeStrToProtoMap() | 27 July 2018, 21:24:09 UTC |
refs/heads/bowbao/onnx_lrn_mvn_layerNorm | 4226f37 | add check for placeholder in cntk c++ LRN. update comment. | 20 July 2018, 19:41:30 UTC |
refs/heads/bowbao/onnx_match_number | fef3029 | merge conflict | 31 October 2018, 17:31:29 UTC |