https://github.com/Microsoft/CNTK

sort by:
Revision Author Date Message Commit Date
ad1f8bc Update conv input space filled by kernel check. 31 July 2018, 22:03:09 UTC
279a466 Squash of the following 1. fix gather and prelu arguments order issue. 2. add typenameToTypeProto map initialization. 3. add CNTK native default lotus logger. 4. add CNTK unsqueeze op. 31 July 2018, 02:58:55 UTC
6ed9659 partially support matmul broadcast. 30 July 2018, 20:29:58 UTC
2964f1f batchnorm_without_batch_axis_fixes 30 July 2018, 19:42:35 UTC
ffc9e14 Merge branch 'sedymche/onnx-reshape' into cntkteam/onnx_without_batch_axis 28 July 2018, 01:43:42 UTC
56adb88 Merge branch 'sedymche/onnx-maxes' into cntkteam/onnx_without_batch_axis 28 July 2018, 00:59:58 UTC
e98b00d Adding support for StopGradient as a no-op. 27 July 2018, 23:21:46 UTC
b7f7b53 Update Flatten/Reshape ONNX conversions for Softmax/Hardmax/LogSoftmax. 27 July 2018, 21:43:52 UTC
6783e9c Fix Hardmax/Softmax/LogSoftmax ONNX import/export. 27 July 2018, 21:37:16 UTC
2122056 Fix ONNX Reshape import when shape constant data is in raw_data field. 27 July 2018, 20:50:52 UTC
8a219b1 Don't crash on non-constant shape on ONNX Reshape import. 26 July 2018, 00:58:12 UTC
d0e68a5 Remove special casing in conv to match length of auto_pad and strides with input dims. Fixes failing onnx backend conv tests. 25 July 2018, 16:19:41 UTC
50d62a6 Merge branch 'sedymche/onnx-backend-test' into cntkteam/onnx_without_batch_axis 24 July 2018, 03:49:04 UTC
4a536b5 Enabling more round trip tests - mean, sum, select. 24 July 2018, 00:31:24 UTC
afd542f Update ONNX backend test skip list. 24 July 2018, 00:06:20 UTC
1af71cf Overhaul of ONNX AveragePool and MaxPool import. Pad op insertion removed. 23 July 2018, 17:00:52 UTC
e2e8d85 disable onnx_op_test layerNormalization float16 due to bug in build test. 21 July 2018, 06:41:36 UTC
dd48e83 lrn, mvn, ln + remove AdjustBroadcastShape update support for LocalResponseNormalization, MeanVarianceNormalization and LayerNormalization 20 July 2018, 22:52:01 UTC
7f59333 Update support for MatMul and Gemm DEBUG: helper prints. add more matmul tests. Add comment for the mysterious swap of input position of ONNXToCNTK Times operator. support import/export on matmul. add direct export of equivalent block ops. update Operator adding ConvertedBlockOP DEBUG: remove prints update comments update dim_denotation for verify_two_input update import/export(block) for Gemm DEBUG: remove prints move Gemm to internal move MatMul to internal. update PrimitiveFunction::Attributes for Gemm. small changes on nit and param name etc. update PrimitiveFunction attribute names. typo update documentation for arguments order change 20 July 2018, 22:52:01 UTC
1eb5183 Fix arguments interface to respect pythonOperandOrder. 20 July 2018, 22:52:01 UTC
fae5dc6 Merge branch 'master' into cntkteam/onnx_without_batch_axis 20 July 2018, 21:55:56 UTC
91d9901 Squashed commit of the following: commit a08e22e7b812891dd641cc512198528a3cd1cddc Author: TJ <tix@microsoft.com> Date: Thu Jul 19 17:53:21 2018 -0700 Enabled example tests for linux commit 2b0b7c49f7ab3ad8ba40684efdaf4c4e4fd2265f Author: Tix <tix@microsoft.com> Date: Thu Jun 28 18:30:51 2018 -0700 Enabled V2 lib CS unit tests on linux 20 July 2018, 16:33:25 UTC
ad63107 Reduction ops Fixes: shape mistmatch and default axes 19 July 2018, 17:03:09 UTC
1c7ebd2 Fix onnx_op_test extra round trip testing code. 19 July 2018, 16:33:51 UTC
03c7c93 Fixing DepthToSpace and SpaceToDepth import/export to enable round trip test. 18 July 2018, 20:52:14 UTC
3806ec6 Change to more numerically stable LogSoftMax implementation. 18 July 2018, 17:31:41 UTC
af3182c Merge remote-tracking branch 'origin/master' into cntkteam/onnx_without_batch_axis 18 July 2018, 17:26:17 UTC
c15c244 Updating ConvTranspose to pass roundtrip + updating onnx_op_test to do one more round of saving. 18 July 2018, 00:42:23 UTC
12e50f2 Removing unncessary print statement in SequentialConvolution layer. 17 July 2018, 23:08:21 UTC
64deb5c Removed references to files that no longer exist. (cherry picked from commit aa13eec9f29b45aea97a7f9e26519cf73a936216) 17 July 2018, 21:39:52 UTC
74fda28 Generated new code signing keys that expire on 12/31/2039 (cherry picked from commit d58886f8d42def3a6bbe263a676c134817bb8b4e) 17 July 2018, 21:39:25 UTC
d69fd95 Adding input validation for group convolution. 17 July 2018, 20:57:00 UTC
aa13eec Removed references to files that no longer exist. 16 July 2018, 21:57:16 UTC
d58886f Generated new code signing keys that expire on 12/31/2039 16 July 2018, 21:52:50 UTC
5ce390c Adding automatic pack/unpack wrapper for simple batch axis ONNX ops. MaxPool, AveragePool, GlobalAveragePool, GlobalMaxPool enabled. 16 July 2018, 20:59:29 UTC
fb28ccf disabling failing tests after changes to support no batch axis mode 16 July 2018, 19:48:40 UTC
09dcac3 Fixing test code in onnx_op_test.py to accurately capture CNTK model shape, based on batch/no batch axis ops. 16 July 2018, 18:48:18 UTC
6e42100 Correcting the model shape value that is compared in the test. 13 July 2018, 21:24:20 UTC
1816043 Update to onnx_op_test code to fix extra (1,) dimension issue. 13 July 2018, 20:58:28 UTC
749abfb First step in removing batch axis generation in ONNX import. Conv import/export works with static axes only. 12 July 2018, 21:21:24 UTC
9ce967a Support one input case for Min/Max to pass ONNX backend tests. 12 July 2018, 17:43:26 UTC
dca867c add warning for convolution padding on channel axis. 11 July 2018, 23:03:56 UTC
5ca4bb3 Add Sequential Convolution. adding convolution over sequential axis related tests. adding convolution over sequential axis. currently additional supported parameters: auto padding strides groups support for dilation needs to be tested on GPU. updating PrimitiveOpType SerializationTests that is missing from other commits .. convert tabs to spaces. Refine cpp convolution unit tests. Add dilation tests to python convolution unit tests. more detailed comments on shape change for 1d seq conv with reduction rank 0. And other minor tweaks. add EndToEndTests of sequential convolution on MNIST add init_bias tests for seq conv minor change in comments rename ConvolutionOverSequenceAxisNode. Add comment on cudnn failed new test. add more comments, trim spaces add more comments, remove magic number, add more boundary checks. remove the last SetValue for outputSeqAxisDimValue as TensorView Unary Op has already updated the value. fix bug in python seqconv default bias shape, and add related unit tests. small tweak in seq conv to avoid additional gpu memory allocation and increase performance. Example: seq MNIST, and profiling adjust conv c++ value unit test channel size. small update on python seq mnist Sequential convolution v2. * re-designed ConvolutionSequenceShapeNode: refactored to separate out computing output sequence length from v1 node design. And refactored ConvolutionNodeBaseExtended as their common base class. (Since "ConvolutionNodeBase" is not only base class of ConvolutionNode but also PoolingNode). * Performance increase against v1. - compute sequence length by MBLayout instead of mask output from unpack. Avoiding the unnecessary cpu/gpu memory copy. not include py sequence example for now .. need to find they a correct location. add check for truncated sequences in sequential convolution improve code style. Moving sequential convolution in python to a new high level api, to maintain compatibility with previous implementation (special case 1d sequential convolution). Add ConvolutionSequenceShape OP. nit update conv_attribute test for updated convolution parameter move sequential parameter to the last update test shortcircuit for CPU convolution dilation. update endtoendtest - unittest baseline file for new convolution unittests. update makefile to include new unittest file for linux nit Update ConvolutionNode initialization code to handle TransformerNode Initialization. nit nit 11 July 2018, 04:10:33 UTC
8bd0459 CPU device should not try to initialize nvml in NcclComm 10 July 2018, 17:50:38 UTC
5aa511d Project updates to use a consistent native runtime 10 July 2018, 15:29:11 UTC
1f0eed8 Merge branch 'liqun/gslfixstage' 10 July 2018, 01:07:41 UTC
83370c1 Squashed commit of the following: commit 02eb64bf5e9f6c22138b5111f5518f6087cea7e0 Author: TJ <tix@microsoft.com> Date: Mon Jul 9 13:07:13 2018 -0700 set the multiverso lib file when asgd is set to true, otherwise it will look for a lib that doesn't match any rule when asgd=false 09 July 2018, 23:41:26 UTC
3c87d20 upgrade linux build from c++11 to c++14, enable gsl, update with latest LotusIR 09 July 2018, 20:29:47 UTC
55b4606 Merge branch 'master' into yanchen/docker-copy 09 July 2018, 17:01:33 UTC
add2896 Not hard-coding build locations 09 July 2018, 15:37:44 UTC
e8f2db4 Add bool in/out support to pass ONNX backend tests. 07 July 2018, 04:50:53 UTC
624bf7d String changes to support conversion from ASCII, UCS2, UCS4, UTF8, UTF16, and UTF32 strings 07 July 2018, 01:48:27 UTC
ed9de47 Merge branch 'yanchen/mpi-makefile' 07 July 2018, 00:54:06 UTC
c11f51d Merge branch 'sedymche/onnx-backend' 07 July 2018, 00:29:50 UTC
2626ae7 Merge branch 'master' into yanchen/mpi-makefile 06 July 2018, 21:36:58 UTC
70cf3a3 Merge branch 'master' into yanchen/docker-copy 06 July 2018, 21:36:37 UTC
7b3a840 Update skip list for ONNX backend tests. 06 July 2018, 20:20:39 UTC
afbe91e update NoGPU.cpp 06 July 2018, 18:08:36 UTC
19ffa06 Merge branch 'master' into bowbao/gather_grad 06 July 2018, 17:03:13 UTC
6adce79 Merge branch 'master' into bowbao/gather_grad 06 July 2018, 17:01:44 UTC
10c1d6b nit 06 July 2018, 16:54:26 UTC
3bfa12b Merge branch 'master' of https://github.com/Microsoft/CNTK into tix/fix_netcoreTestFailure 06 July 2018, 04:49:20 UTC
0b317eb Merge branch 'bowbao/clone_segfault' 06 July 2018, 02:04:49 UTC
92796cf fixed bug around mapping from columnValidMask to data. Instead of a 1 to 1 mapping, each element in columnValidMask maps to a grid of multiple rows and columns in data matrix. 06 July 2018, 01:37:40 UTC
04844b0 Fixed path 06 July 2018, 01:22:25 UTC
610abaf Added comment 05 July 2018, 23:59:38 UTC
a89a334 Fixed path 05 July 2018, 23:55:33 UTC
dc29eb3 Enabled the test for py35 so that it gets run by default in bvt 05 July 2018, 23:52:10 UTC
6cc772f small change for more reuse of code 05 July 2018, 23:42:51 UTC
56d0ada remove the data folder created by the python script 05 July 2018, 22:46:41 UTC
6e265da Fixed an issue for copying patches Unlike Linux-like cp instruction, docker COPY instruction doesn't preserve the top-level dir structure (e.g. it copies the contents of the top-level dir rather than the dir itelse to the image). Our Dockerfile for generating GPU image has correct behavior though. 05 July 2018, 21:42:02 UTC
9a075b3 Merge branch 'master' into bowbao/clone_segfault 05 July 2018, 21:40:44 UTC
0df5c39 Add method ScatterToIndicesWithMask for class Matrix(both CPUMatrix and GPUMatrix). Change GatherNode Backprop to use the new method instead of altering data for gaps in input matrix. 05 July 2018, 21:20:13 UTC
49e3583 Copy onnx_cntk from https://github.com/onnx/onnx-cntk. 05 July 2018, 21:13:40 UTC
9667426 Fixing run-test to grab for the correct dll for netcore build 05 July 2018, 19:12:36 UTC
ed9f879 Merge branch 'master' into yanchen/mpi-makefile 05 July 2018, 17:00:01 UTC
17e6a01 Merge branch 'pull/3277' 04 July 2018, 03:49:50 UTC
7c74244 set missing value back to 0 in case of unexpected usage. 04 July 2018, 01:48:01 UTC
1e058ce fix Gather op's incorrect gradient value. * the error was due to that we pad 0 as default value for missing gaps. All these then each contribute 1 to the gradient of reference at index 0. The fix is to mask missing values in indices matrix to negative, and in Matrix scatter implementation to check and skip negative indices. (previous Matrix CPU implementation already checks for negative indices) 04 July 2018, 01:34:11 UTC
c049493 Merge branch 'master' into yanchen/mpi-makefile 03 July 2018, 23:07:22 UTC
81ced59 add None check for python clone substitution. * it seems None somehow escapes the typecheck when being converted to class CNTK::Variable thus causing the crash. 03 July 2018, 22:19:33 UTC
8577222 Merge branch 'tix/convertEvalExampleTestToNetCore' of https://github.com/Microsoft/CNTK 03 July 2018, 22:02:22 UTC
25a0896 use system CXX for building Multiverso on Linux We used to set CXX to mpic++, which was then used for building Multiverso. Unfortunately, this kind of configuration conflicted with find_package(MPI), and hence caused failure. This commit fixed the issue by using system CXX for building Multiverso. It also fixed two other issues: * EVAL_LIB depends on libmultiverso, so we had to set MULTIVERSO_LIB before the EVAL_LIB rule * cmake variable doesn't have a "PATHNAME" type, we used PATH instead. We also replace FILEPATH with PATH because BOOST_LIBRARY_DIRS points to a path rather than a file. 03 July 2018, 20:55:57 UTC
dd9705b Fixing LRN op to match ONNX 1.2 spec. Size attribute has semantics of diameter and not radius, as was the case before. 03 July 2018, 19:40:12 UTC
8d39322 Converted eval example to use netcore 03 July 2018, 01:33:08 UTC
63603e0 SqueezeNet numbers mistmatch fix (due to an issue on softmax layer) 02 July 2018, 20:58:49 UTC
07fd96f support ONNX1.2.2 cast op 02 July 2018, 15:38:50 UTC
2f91f74 Fix saved model name output Corrected the output showing the name of the model being saved 01 July 2018, 13:53:29 UTC
bece037 Merge branch 'sedymche/onnx-trig' 29 June 2018, 01:52:40 UTC
69938f2 Adding full support for 'alpha' attribute in ELU op in ONNX and CNTK. 28 June 2018, 22:34:30 UTC
e95f92c Add Tan/Atan ops to CNTK (with ONNX support). 28 June 2018, 21:03:02 UTC
c589619 update to ONNX1.2.2 28 June 2018, 15:38:36 UTC
e160304 Merge branch 'sptiwari/onnx_avg_pool_stride' 28 June 2018, 04:02:28 UTC
6340792 Merge branch 'zhujie/straightthrough_unaryop' 28 June 2018, 00:55:55 UTC
2e29625 Merge branch 'tix_fixRestoredTestFailures' of https://github.com/Microsoft/CNTK 27 June 2018, 22:25:50 UTC
805c4a6 Fixing strides and dilation attributes in ONNX AveragePool, MaxPool, Conv, ConvTranpose import. 27 June 2018, 21:54:47 UTC
3c389ed fixing a bug in straight through implementation 27 June 2018, 18:44:49 UTC
cc368ae fixing a bug in backwarod propagation of straightthrough 27 June 2018, 18:44:49 UTC
8c1e5ed fixing type mismatch in tensorops.h and changing name of python name for straightthrough 27 June 2018, 18:44:49 UTC
02be8a0 adding straight through unary op 27 June 2018, 18:44:49 UTC
back to top