platform/upstream/dldt.git
4 years ago[GNA] Added support permute layer (#723)
Andrey Dmitriev [Tue, 9 Jun 2020 13:43:01 +0000 (16:43 +0300)]
[GNA] Added support permute layer (#723)

* [GNA] Added GNA natively supported permute layer cases.

4 years ago[IE VPU] Dynamic Broadcast tests (#737)
Maksim Doronin [Tue, 9 Jun 2020 13:10:12 +0000 (16:10 +0300)]
[IE VPU] Dynamic Broadcast tests (#737)

* [IE VPU] Enable StaticShapeBroadcast tests

* [IE VPU] DSR: support case when shape is output and input for stage

* [IE VPU] Enable Broadcast and Transpose tests

* [IE VPU] DSR: fix typo

* [IE VPU] Add assertion for numConsumer in DSR

* [IE VPU] Added CheckMyriadX helper method

* [IE VPU] New DSR assert for input->getInputTo

* [IE VPU] Fix myriad2 tests bug

4 years agoConst folding and reference implementation for PriorBox(Clustered) ops (#785)
Ivan Tikhonov [Tue, 9 Jun 2020 11:47:49 +0000 (14:47 +0300)]
Const folding and reference implementation for PriorBox(Clustered) ops (#785)

* Constant folding for PriorBox, PriorBoxClustered; Deleted PriorBoxIE, PriorBoxClusteredIE and transformations; Added unit tests; codestyle

* Delete debug info

* delete unnecessary convert_prior_to_ie_prior.hpp file

* fix ngraph reader tests; delete PriorBoxIE functional test

* fix for ngraph reader tests

* Apply review comment

* apply ngraph codestyle

* restore PriorBoxClustered tests in disabled state

4 years agoadd support for multiple scale factors in speech sample (#835)
Denis Orlov [Tue, 9 Jun 2020 11:36:28 +0000 (14:36 +0300)]
add support for multiple scale factors in speech sample (#835)

Co-authored-by: Anna Alberska <anna.alberska@intel.com>
4 years ago[IE VPU] Update firmware; enable convolution VPU OCL tests (#802)
Andrey Sokolov [Tue, 9 Jun 2020 11:34:10 +0000 (14:34 +0300)]
[IE VPU] Update firmware; enable convolution VPU OCL tests (#802)

4 years ago[IE CLDNN] Added 6d tensor support in eltwise/scale primitives (#826)
Vladimir Paramuzov [Tue, 9 Jun 2020 11:29:36 +0000 (14:29 +0300)]
[IE CLDNN] Added 6d tensor support in eltwise/scale primitives (#826)

4 years ago[ci-skip][IE MKLDNN] Add Precision U16 in MKLDNN (#783)
Anton Zaytsev [Tue, 9 Jun 2020 11:20:43 +0000 (14:20 +0300)]
[ci-skip][IE MKLDNN] Add Precision U16 in MKLDNN (#783)

4 years ago[TESTS] fixes after rebase
dmitrygo [Tue, 9 Jun 2020 08:00:10 +0000 (11:00 +0300)]
[TESTS] fixes after rebase

4 years ago[CPU] mkldnn submodule up
dmitrygo [Tue, 9 Jun 2020 07:28:20 +0000 (10:28 +0300)]
[CPU] mkldnn submodule up

4 years ago[LPT] Fixed quantizeBlob routine for 3D case
dmitrygo [Tue, 9 Jun 2020 06:37:03 +0000 (09:37 +0300)]
[LPT] Fixed quantizeBlob routine for 3D case

4 years ago[LPT] GPU tests were fixed
Edward Shogulin [Tue, 9 Jun 2020 03:59:37 +0000 (06:59 +0300)]
[LPT] GPU tests were fixed

4 years ago[LPT] [TEST] LayerTransformation test threshold was updated
Edward Shogulin [Mon, 8 Jun 2020 19:59:12 +0000 (22:59 +0300)]
[LPT] [TEST] LayerTransformation test threshold was updated

4 years ago[CPU] Gemm node: supported precisions U8 and I8 and added tests
Anton Voronov [Wed, 3 Jun 2020 17:01:58 +0000 (20:01 +0300)]
[CPU] Gemm node: supported precisions U8 and I8 and added tests

4 years ago[LPT] FullyConnected & Gemm tests
Edward Shogulin [Mon, 8 Jun 2020 14:15:45 +0000 (17:15 +0300)]
[LPT] FullyConnected & Gemm tests

4 years ago[CPU] Fixed weights candidate initialization in FC node
dmitrygo [Thu, 4 Jun 2020 17:13:59 +0000 (20:13 +0300)]
[CPU] Fixed weights candidate initialization in FC node

4 years ago[LPT] Gemm and FullyConnected 3D improvement
Edward Shogulin [Thu, 4 Jun 2020 10:24:18 +0000 (13:24 +0300)]
[LPT] Gemm and FullyConnected 3D improvement

4 years ago[LPT] tests extending
Edward Shogulin [Mon, 1 Jun 2020 07:37:39 +0000 (10:37 +0300)]
[LPT] tests extending

4 years ago[LPT] [Test] Low precision transformations functional tests infrastructure improvement
Edward Shogulin [Sun, 31 May 2020 19:14:54 +0000 (22:14 +0300)]
[LPT] [Test] Low precision transformations functional tests infrastructure improvement

4 years ago[LPT] FullyConnected transformation fix
Edward Shogulin [Sat, 30 May 2020 12:01:28 +0000 (15:01 +0300)]
[LPT] FullyConnected transformation fix

4 years ago[nGraph] Remove Reshape for 3D FullyConnected
Edward Shogulin [Sat, 30 May 2020 11:31:23 +0000 (14:31 +0300)]
[nGraph] Remove Reshape for 3D FullyConnected

4 years ago[IE COMMON] [LPT] Concat asymmetric quantization with signed interval fix
Edward Shogulin [Sat, 30 May 2020 10:56:01 +0000 (13:56 +0300)]
[IE COMMON] [LPT] Concat asymmetric quantization with signed interval fix

4 years ago[IE COMMON] [LPT] Support 3D layout for FullyConnected transformation
Edward Shogulin [Sat, 30 May 2020 09:57:36 +0000 (12:57 +0300)]
[IE COMMON] [LPT] Support 3D layout for FullyConnected transformation

4 years ago[CPU] Added custom implementations (power=0.5, power=-1.0) for Power node
dmitry-gorokhov [Fri, 29 May 2020 13:56:20 +0000 (16:56 +0300)]
[CPU] Added custom implementations (power=0.5, power=-1.0) for Power node

4 years ago[CPU][WA] Supported 3D layout for FullyConnected primitive
dmitry-gorokhov [Fri, 29 May 2020 13:53:59 +0000 (16:53 +0300)]
[CPU][WA] Supported 3D layout for FullyConnected primitive

Extended jit uni depthwise primitive to support 3D inputs

4 years ago[Common] Static analyzed issues. (#804)
Nikolay Shchegolev [Tue, 9 Jun 2020 10:49:50 +0000 (13:49 +0300)]
[Common] Static analyzed issues. (#804)

4 years ago[IE CLDNN] Better error message when output is not found (#824)
Vladimir Paramuzov [Tue, 9 Jun 2020 09:26:28 +0000 (12:26 +0300)]
[IE CLDNN] Better error message when output is not found (#824)

4 years ago[LPT] [TEST] LayerTransformation test threshold was updated (#828)
Edward Shogulin [Tue, 9 Jun 2020 07:34:17 +0000 (10:34 +0300)]
[LPT] [TEST] LayerTransformation test threshold was updated (#828)

4 years ago[CPU] fix set up config for bin conv fused (#608)
Maxim Andronov [Tue, 9 Jun 2020 06:59:29 +0000 (09:59 +0300)]
[CPU] fix set up config for bin conv fused (#608)

4 years ago[Python API] Fixate requirements versions (#830)
Anastasia Kuporosova [Tue, 9 Jun 2020 05:49:49 +0000 (08:49 +0300)]
[Python API] Fixate requirements versions (#830)

4 years ago[Python API] Fix deprecation warnings (#812)
Anastasia Kuporosova [Tue, 9 Jun 2020 05:48:08 +0000 (08:48 +0300)]
[Python API] Fix deprecation warnings (#812)

4 years ago[LPT] [CPU] NormalizeL2 transformation (#662)
Edward Shogulin [Mon, 8 Jun 2020 19:42:50 +0000 (22:42 +0300)]
[LPT] [CPU] NormalizeL2 transformation (#662)

* [LPT] NormalizeL2 transformation

* [LPT] NormalizeL2 transformation tests improvement

* [CPU] Fixed depthwise injector aux_vec_count for broadcasting case

* [LPT] Normalize on GPU enabling

Co-authored-by: Zinoviev, Vladimir <vladimir.zinoviev@intel.com>
Co-authored-by: dmitrygo <dmitry.gorokhov@intel.com>
4 years agoAdd fixedscale(bias) components to Kaldi (#725)
Pavel Esir [Mon, 8 Jun 2020 18:37:44 +0000 (21:37 +0300)]
Add fixedscale(bias) components to Kaldi (#725)

* Added fixed scale(bias) components

* Successfully converted after adding fixed bias,scale components

* Added unittests

4 years ago[IE CLDNN] Fix variable initialization issues (#816)
Mikhail Letavin [Mon, 8 Jun 2020 18:07:50 +0000 (21:07 +0300)]
[IE CLDNN] Fix variable initialization issues (#816)

4 years ago[ICV] Watchdog switch + ddr initialization (#554)
Nikita Kudriavtsev [Mon, 8 Jun 2020 17:51:45 +0000 (20:51 +0300)]
[ICV] Watchdog switch + ddr initialization (#554)

* [IE Myriad] Added XLinkBootFirmware method in XLink API for booting firmware buffer

* [IE Myriad] Patch firmware in mvnc. Added test to check device reset without connecting.

* [IE Myriad] Added option MOVIDIUS_DDR_TYPE for Myriad plugin

* [IE Myriad] Added tests for new option MOVIDIUS_DDR_TYPE

* [IE Myriad] Update firmware 1201 -> 1212

* [IE Myriad] Convolution3x3 tests are disabled due to firmware issue. #-32921

4 years ago[IE TEST] Eltwise tests refactoring (#726)
Irina Efode [Mon, 8 Jun 2020 15:44:42 +0000 (18:44 +0300)]
[IE TEST] Eltwise tests refactoring (#726)

* [IE TEST] Eltwise tests refactoring

* [IE TESTS] Fix comments

4 years ago[GNA] Support changing the execution mode in runtime (#801)
Denis Orlov [Mon, 8 Jun 2020 15:43:12 +0000 (18:43 +0300)]
[GNA] Support changing the execution mode in runtime (#801)

4 years agoFix unit tests for select layer. (#638)
iliya mironov [Mon, 8 Jun 2020 15:39:40 +0000 (18:39 +0300)]
Fix unit tests for select layer. (#638)

* Fix unit tests for select layer.

4 years agoImplement support for opset3 EmbeddingBag ops (#546)
Maxim Vafin [Mon, 8 Jun 2020 15:06:40 +0000 (18:06 +0300)]
Implement support for opset3 EmbeddingBag ops (#546)

* [MO] Implement EmbeddingBag_3

* Transform dynamic sub-graph of Wide and Deep into EmbeddingSegmentsSum

- Expressed SparseWeightedSum sub-graph through EmbeddingSegmentsSum
- Removed experimental SparseWeightedSum layer
- Implemented tests for the transformation

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix EmbeddingBag shape infer

* Fix EmbeddingSegmentsSum transformation for Wide and Deep

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix EmbeddingSegmentSum replacer after ports swap

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Update package_BOM.txt

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Add unit tests for EmbeddingXXX shape infer

* Fix ATen resolver

* Remove deleted files from BOM

* Add opset version to embedding_bag

* Use base class for EmbeddingBag

* Fix per_sample_weights case

* Fix EmbeddingSegmentsSum transformation

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix EmbeddingBag checks

* Fix ATen front transformation and merge conflicts

* Fix BOM

* Work around limitation for I64 input of W&D model

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Cleanup where operation to fix affect of WhereDecomposition transform

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix BOM

* Correct EmbeddingSegmentSum transform for Wide and Deep

Add casting segment ids to i32 and remove ConstToResult sub-graph.

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Update BOM with RemoveConstToResult transform

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Add more comments for RemoveConstToResult transformation

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Remove useless logging in EmbeddingSegmentsSum transformation

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Small fixes

* Move EmbeddingBag resolving back to front phase

* Improve error messages

* Fix typo in unittests

* Reimplement sparse_reshape middle transform

Avoid deprecated API.

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Clean-up graph after sparse_reshape and ConstToResult transformation

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix clean-up for transformations

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix clean-up for transformation #2

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
Co-authored-by: Roman Kazantsev <roman.kazantsev@intel.com>
4 years ago[IE CLDNN] Optimize 1x1 imad convolution kernel (#757)
Konrad Dobros [Mon, 8 Jun 2020 13:44:50 +0000 (15:44 +0200)]
[IE CLDNN] Optimize 1x1 imad convolution kernel (#757)

4 years agoAdd commit links to memcheck report (#820)
Andrey Somsikov [Mon, 8 Jun 2020 12:08:58 +0000 (15:08 +0300)]
Add commit links to memcheck report (#820)

4 years agoDisable warnings-as-errors for ONNX target (#749)
Tomasz Dołbniak [Mon, 8 Jun 2020 10:52:45 +0000 (12:52 +0200)]
Disable warnings-as-errors for ONNX target (#749)

* Disable warnings-as-errors for ONNX target

* Disable warnigs-as-errors for windows too

* Change WIN32 -> MSVC

4 years ago[ MO ] Groupped conv fusion (#797)
Evgenya Stepyreva [Mon, 8 Jun 2020 10:00:54 +0000 (13:00 +0300)]
[ MO ] Groupped conv fusion (#797)

Fixed the group convolution fusion pass to properly get the feature dim in NCHW layout case.

4 years ago[LPT] [Test] DepthToSpace sporadic fail fix (#815)
Edward Shogulin [Mon, 8 Jun 2020 09:55:37 +0000 (12:55 +0300)]
[LPT] [Test] DepthToSpace sporadic fail fix (#815)

4 years agoAzure CI: increase timeout for Mac to 180 min
Alexander Zhogov [Mon, 8 Jun 2020 09:17:50 +0000 (12:17 +0300)]
Azure CI: increase timeout for Mac to 180 min

4 years ago[IE CLDNN] Enable ShuffleChannels op (#787)
Vladimir Paramuzov [Sun, 7 Jun 2020 19:57:20 +0000 (22:57 +0300)]
[IE CLDNN] Enable ShuffleChannels op (#787)

4 years ago[LPT] [CPU] DepthToSpace transformation (#663)
Edward Shogulin [Sun, 7 Jun 2020 18:12:52 +0000 (21:12 +0300)]
[LPT] [CPU] DepthToSpace transformation (#663)

* [LPT] [TEST] LayerTransformation generalization

* [LPT] DequantizationDetails extending

* [LPT] DepthToSpace transformation implementation

4 years agoAzure: Add Ninja (#803)
Alexander Zhogov [Sat, 6 Jun 2020 12:56:24 +0000 (15:56 +0300)]
Azure: Add Ninja (#803)

* Azure: Add Ninja

* Fix 'Install Ninja' on Linux

* Fix bin dir path on Windows

* Add -Wno-unused-variable on Mac

* Add -Wno-error=unused-command-line-argument on Mac

* Set CXXFLAGS for Mac

* Improvements

* Fix BIN_DIR on Linux

4 years ago[IE CLDNN] Enable DepthToSpace (#780)
Vladimir Paramuzov [Fri, 5 Jun 2020 17:16:47 +0000 (20:16 +0300)]
[IE CLDNN] Enable DepthToSpace (#780)

Enabled DepthToSpace ngraph transformat
Updated implementation to support 5d and mode parameter
fsv16 direct support
Functional tests for GPU

4 years agoit is duplicate of PR: #656, but without test modification (#794)
Daria Mityagina [Fri, 5 Jun 2020 16:57:03 +0000 (19:57 +0300)]
it is duplicate of PR: #656, but without test modification (#794)

4 years agoDisable Hw Avg Pooling for small output tensors if excludePad=true (#772)
Chance Luo [Fri, 5 Jun 2020 16:47:53 +0000 (00:47 +0800)]
Disable Hw Avg Pooling for small output tensors if excludePad=true (#772)

4 years agoadd /wd4819 to disable C4819 warning, which is treated as error in win32 (#767)
Kami-996 [Fri, 5 Jun 2020 13:04:59 +0000 (21:04 +0800)]
add /wd4819 to disable C4819 warning, which is treated as error in win32 (#767)

Co-authored-by: jasonlee <jasonlee@qiyi.com>
4 years agoFix: ITT_INCLUDE_DIR was not correctly detected (#748)
emmanuelattia-philips [Fri, 5 Jun 2020 11:46:39 +0000 (13:46 +0200)]
Fix: ITT_INCLUDE_DIR was not correctly detected (#748)

4 years ago[IE CLDNN] GEMM int8 optimization using MMAD macro (#635)
Ilya Znamenskiy [Fri, 5 Jun 2020 11:28:21 +0000 (14:28 +0300)]
[IE CLDNN] GEMM int8 optimization using MMAD macro (#635)

4 years ago[CPU] supported ShuffleChannels and added tests (#636)
Anton Voronov [Fri, 5 Jun 2020 11:10:55 +0000 (14:10 +0300)]
[CPU] supported ShuffleChannels and added tests (#636)

4 years agoSave the name of output data if we remove previous layer (#760)
Ilya Churaev [Fri, 5 Jun 2020 10:36:35 +0000 (13:36 +0300)]
Save the name of output data if we remove previous layer (#760)

* Save the name of output data if we remove previous layer

* Added test

4 years ago[GNA] Set default GNA library to GNA2 (#771)
Pavel Rodionov [Fri, 5 Jun 2020 10:00:58 +0000 (13:00 +0300)]
[GNA] Set default GNA library to GNA2 (#771)

4 years ago[CPU] DepthToSpace review leftovers
dmitrygo [Thu, 4 Jun 2020 12:59:14 +0000 (15:59 +0300)]
[CPU] DepthToSpace review leftovers

4 years agoAdd ReduceL2 decomposition (#733)
Maxim Vafin [Fri, 5 Jun 2020 09:34:57 +0000 (12:34 +0300)]
Add ReduceL2 decomposition (#733)

* Add ReduceL2 decomposition

* Add ReduceL2 transformation tests

* Add const propagation unit test for ReduceL2

4 years agoFix hetero mode in speech sample - set config when loading network (#786)
Denis Orlov [Fri, 5 Jun 2020 08:54:03 +0000 (11:54 +0300)]
Fix hetero mode in speech sample - set config when loading network (#786)

4 years ago[SAMPLES] Use defined constant instead of string (#788)
Ilya Churaev [Fri, 5 Jun 2020 08:22:24 +0000 (11:22 +0300)]
[SAMPLES] Use defined constant instead of string (#788)

4 years ago[IE][VPU][nGraph]: Fixes DTS transformations to properly keep outputs names (#734)
Gladilov, Gleb [Fri, 5 Jun 2020 08:16:52 +0000 (11:16 +0300)]
[IE][VPU][nGraph]: Fixes DTS transformations to properly keep outputs names (#734)

* NonZero, Broadcast

* Concat

* Gather

* [IE][VPU][nGraph]: Fixes DTS transformations to correctly keep outputs names

* [IE][VPU][nGraph]: Fixes dynamic to static shape nonzero tests

Co-authored-by: Roman Vyunov <roman.vyunov@intel.com>
4 years ago[LPT] Output layers update fix (#754)
Edward Shogulin [Fri, 5 Jun 2020 07:54:38 +0000 (10:54 +0300)]
[LPT] Output layers update fix (#754)

4 years ago[IE CLDNN] Improve Gather performance and add fusing support (#736)
Sergey Shlyapnikov [Fri, 5 Jun 2020 07:20:58 +0000 (10:20 +0300)]
[IE CLDNN] Improve Gather performance and add fusing support (#736)

4 years ago[IE CLDNN] Implement NormalizeL2 int8 kernels (#720)
Egor Churaev [Fri, 5 Jun 2020 07:16:27 +0000 (10:16 +0300)]
[IE CLDNN] Implement NormalizeL2 int8 kernels (#720)

4 years agoAvoid loading of reader if it doesn't exist (#758)
Ilya Churaev [Thu, 4 Jun 2020 18:21:13 +0000 (21:21 +0300)]
Avoid loading of reader if it doesn't exist (#758)

* Avoid loading of reader if it doesn't exist

* Updated error messages

4 years agoFix Proposal for the case of 2 outputs (#773)
Maxim Vafin [Thu, 4 Jun 2020 17:56:46 +0000 (20:56 +0300)]
Fix Proposal for the case of 2 outputs (#773)

4 years agoAdded default value for 'aligned' in the ExperimentalDetectronROIFeatureExtractor...
Evgeny Lazarev [Thu, 4 Jun 2020 17:47:52 +0000 (20:47 +0300)]
Added default value for 'aligned' in the ExperimentalDetectronROIFeatureExtractor for backward compatibility (#777)

Fixed backward compatibility issue that old IRs with ExperimentalDetectronROIFeatureExtractor operation cannot be loaded with the new IE

4 years ago[IE CLDNN] Permute fused ops support (#642)
Lukasz Debski [Thu, 4 Jun 2020 14:01:21 +0000 (16:01 +0200)]
[IE CLDNN] Permute fused ops support (#642)

4 years ago[CPU][BF16] Default Optimisation Capability of BF16 was enabled on CPX (#647)
Alexey Varyzgin [Thu, 4 Jun 2020 13:06:15 +0000 (16:06 +0300)]
[CPU][BF16] Default Optimisation Capability of BF16 was enabled on CPX (#647)

4 years agoAdd support for ONNX Pad-11 (#744)
Maxim Vafin [Thu, 4 Jun 2020 11:48:31 +0000 (14:48 +0300)]
Add support for ONNX Pad-11 (#744)

4 years ago[GNA] Support 100 inputs, instead of 10 (#741)
Denis Orlov [Thu, 4 Jun 2020 11:33:09 +0000 (14:33 +0300)]
[GNA] Support 100 inputs, instead of 10 (#741)

4 years agoDepthToSpace, SpaceToDepth layers optimizations (#706)
Gorokhov Dmitriy [Thu, 4 Jun 2020 11:25:19 +0000 (14:25 +0300)]
DepthToSpace, SpaceToDepth layers optimizations (#706)

* [CPU] Updated DepthToSpace and SpaceToDepth layers to be conformant with the specification

The patch also includes n[d]hwc layout support as well as some optimizations

* [CPU][TESTS] Removed old DepthToSpace test since it doesn't corresponds to layer's specification

* [nGraph] Utilize CommonOptimizations pass with custom transformations callback

4 years ago[ MO ] InterpolateConcat empty sources fix (#764)
Evgenya Stepyreva [Thu, 4 Jun 2020 11:18:33 +0000 (14:18 +0300)]
[ MO ] InterpolateConcat empty sources fix (#764)

4 years ago[IE CLDNN] Disabled sporadic detection output tests (#740)
Vladimir Paramuzov [Thu, 4 Jun 2020 08:14:05 +0000 (11:14 +0300)]
[IE CLDNN] Disabled sporadic detection output tests (#740)

4 years ago[IE CLDNN] Remove unused fused deps for FQ (#712)
Vladimir Paramuzov [Thu, 4 Jun 2020 07:30:46 +0000 (10:30 +0300)]
[IE CLDNN] Remove unused fused deps for FQ (#712)

Remove unused fused FQ kernel arguments to avoid extra setArg() calls which significantly reduces host overhead

4 years ago[IE CLDNN] Implement EmbeddingBag operations (#623)
Egor Churaev [Thu, 4 Jun 2020 07:25:28 +0000 (10:25 +0300)]
[IE CLDNN] Implement EmbeddingBag operations (#623)

Implemented three operations: EmbeddingBagPackedSum,
EmbeddingBagOffsetsSum and EmbeddingSegmentsSum. These operations do
the same work but have a different format of inputs.

4 years ago[MKLDNN_PLUGIN] Convolution node: skip initializing of primitive descriptors for...
Anton Voronov [Thu, 4 Jun 2020 05:06:14 +0000 (08:06 +0300)]
[MKLDNN_PLUGIN] Convolution node: skip initializing of primitive descriptors for planar layout if there is already jit primitive (#672)

4 years agoRevert "Enabled thread tests (#717)" (#756)
Ilya Lavrenov [Wed, 3 Jun 2020 19:32:55 +0000 (22:32 +0300)]
Revert "Enabled thread tests (#717)" (#756)

This reverts commit 99a2423ec0dc6d4570ca989be97da02376b3e169.

4 years agoMO should support LRN k param with caffe model, rather than fixed to 1 (#716)
wistal [Wed, 3 Jun 2020 17:33:55 +0000 (01:33 +0800)]
MO should support LRN k param with caffe model, rather than fixed to 1 (#716)

Co-authored-by: yipengqu <yipeng.qu@intel.com>
4 years ago[Docs] Fixes in readme files: (#750)
Alexey Suhov [Wed, 3 Jun 2020 17:14:35 +0000 (20:14 +0300)]
[Docs] Fixes in readme files: (#750)

- change repo name to openvino
- update driver version
- fix path to samples data
- remove section about Movidius driver installation
- change latest release to 2020.3
- merge fixes in install_dependencies.sh from 2020 branch

4 years ago[IE CLDNN] Added fusing suport to all pooling kernels (#689)
Mikołaj Życzyński [Wed, 3 Jun 2020 16:44:27 +0000 (18:44 +0200)]
[IE CLDNN] Added fusing suport to all pooling kernels (#689)

adds fusing support to all available pooling kernels
tests all possible input type/output type configurations
fixes minor bug in max pooling in pooling_gpu_test.cpp
fixed minor bug with yxbf format in pooling_gpu_ref and pooling_gpu_int8_ref kernels
fixes bug with b_fs_yx_fsv32 format in pooling_gpu kernel
resolves bug with max pooling accuracy missmatch in case of non zero pad end layer parameter
resolves average pooling accuracy missmatch in case of non zero pad end layer parameter

4 years ago[IE CLDNN] Fixed stack overflow in calculate_prior_boxes pass (#747)
Lukasz Debski [Wed, 3 Jun 2020 16:42:50 +0000 (18:42 +0200)]
[IE CLDNN] Fixed stack overflow in calculate_prior_boxes pass (#747)

The problem behind this error was in program_impl::init_graph() where in calculate_prior_boxes we are trying to calculate output layout of an entire network recursively which causes stack overflow. Calculating output layouts beforehand in processing order fixes this issue.

4 years agoAdd 'aligned' param to ExperimentalDetectronROIFeatureExtractor for CPU plugin and MO
Ilya-Krylov [Wed, 3 Jun 2020 14:52:40 +0000 (17:52 +0300)]
Add 'aligned' param to ExperimentalDetectronROIFeatureExtractor for CPU plugin and MO

4 years agoPython API for Assign, ReadValue and ExtractImagePatches (#719)
Tomasz Dołbniak [Wed, 3 Jun 2020 13:01:43 +0000 (15:01 +0200)]
Python API for Assign, ReadValue and ExtractImagePatches (#719)

4 years agomkldnn_memory_solver.hpp: include stdint.h to avoid build error (#729)
LiweiSong [Wed, 3 Jun 2020 12:19:29 +0000 (20:19 +0800)]
mkldnn_memory_solver.hpp: include stdint.h to avoid build error (#729)

fix the following compile error:

inference-engine/src/mkldnn_plugin/mkldnn_memory_solver.hpp:60:9: error: 'int64_t' does not name a type
|    60 |         int64_t size;
|       |         ^~~~~~~

include stdint.h to fix this.

Signed-off-by: Liwei Song <liwei.song@windriver.com>
4 years ago[LPT] [TEST] Sporadic test fail fix (workaround) (#742)
Edward Shogulin [Wed, 3 Jun 2020 12:05:45 +0000 (15:05 +0300)]
[LPT] [TEST] Sporadic test fail fix (workaround) (#742)

4 years ago[ nGraph ] FP16 for evaluate (#722)
Evgenya Stepyreva [Wed, 3 Jun 2020 11:14:59 +0000 (14:14 +0300)]
[ nGraph ] FP16 for evaluate (#722)

4 years agoUse pre-defined DB collection names in memcheck_upload.py CLI (#651)
Vitaliy Urusovskij [Wed, 3 Jun 2020 10:54:38 +0000 (13:54 +0300)]
Use pre-defined DB collection names in memcheck_upload.py CLI (#651)

Use argparses `choices` for `--db_collection` option.

Also removed unnecessary redefinition of `db_collection` in memcheck_upload.py

4 years ago[IE CLDNN] Improve kernel selection for b_fs_yx_fsv16 layout and optimize Convolution...
Sergey Shlyapnikov [Wed, 3 Jun 2020 10:42:15 +0000 (13:42 +0300)]
[IE CLDNN] Improve kernel selection for b_fs_yx_fsv16 layout and optimize Convolution kernels (#730)

4 years ago[IE TESTS] Move InferRequestTests (#618)
Anton Zaytsev [Wed, 3 Jun 2020 09:16:00 +0000 (12:16 +0300)]
[IE TESTS] Move InferRequestTests (#618)

* [IE TESTS] move Infer_request tests

* fix v0

* [ci-skip][IE TESTS] test update basic class v0

* [ci-skip][IE TESTS] test update basic class v1

* [ci-skip][IE TESTS] test update basic class

* [ci-skip][IE TESTS] test update basic class v3

* [ci-skip][IE TESTS] test update basic class final versions

* [ci-skip][IE TESTS] fix

* [ci-skip][IE TESTS] fix codestaly and comment

Co-authored-by: Irina Efode <irina.efode@intel.com>
4 years agoSkip some functional tests on VPU (#568)
Evgeny Talanin [Wed, 3 Jun 2020 09:15:06 +0000 (12:15 +0300)]
Skip some functional tests on VPU (#568)

4 years ago[ONNX] GRU and RNN operators. (#607)
Adam Osewski [Wed, 3 Jun 2020 09:01:56 +0000 (11:01 +0200)]
[ONNX] GRU and RNN operators. (#607)

* Create generic RecurrentSequenceDirection enum.

* Helper class RecurrentSequenceOp.

* Add ONNX GRU & RNN operators.

* Use OutputVector.

* Update doc.

* Add UTs for GRU and skip them on IE_CPU

* Add UT for bidirectional mode and fix it.

* Normalize activation function name case.

* Add unit-tests for RNN operator.

* UT for GRU with linear_before_reset set to true.

* Fix ONNX GRU for linear_before_reset case.

* Remove unnecessary symbol export macro.

* Fix CentOS error.

* Update UTs.

- Update few tests accuracy tolerance
- Update rnn_fwd_activations with new reference values and model.

* Review comment: add check for static shape

* Add UT for RNN with constant inputs W, R.

* Skip UT with const W,R on IE_CPU

4 years ago[IE][VPU]: Faster-RCNN fixes on myriad plugin side (#711)
Gladilov, Gleb [Wed, 3 Jun 2020 08:43:19 +0000 (11:43 +0300)]
[IE][VPU]: Faster-RCNN fixes on myriad plugin side (#711)

* [IE][VPU]: Enables pass for propagating dynamism to network outputs

If network had dynamic output and then myriad Front-End inserted
convert stage at the end (to convert FP16 -> FP32 - output precision)
then dynamism would not be propagated - we have convert stage that
has dynamic input, but static output. As a result, we have run-time
error in Convert kernel: input and output shapes do not match.

At the moment, pass supports only Convert stage as output stage
over which we should propagate dynamism to outputs.

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>
* [IE][VPU]: Fixes parse DSR in case of output data

Replacing stage output must be done after replacing
data to shape parent, because the last one may access
original parent producer, but after replacing stage output
it'd not have one.

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>
* [IE][VPU]: Fixes MacOS build

* [IE][VPU]: Fixes shape data naming convention

Plugin part assumes that if there is dynamic data object, that's
represented as 2 different data objects (data and shape), then
shape data object has name = data object name + @shape suffix.

Pass that creates new dynamic data object should respect that
assumption.

* [IE][VPU]: Fixes dis-alignment in names of data objects representing dynamic data object

MyriadInferRequest::GetResult assumes that in case of dynamic data object
"data" data object and "shape" data object will have aligned names:
"shape" name = "data" name + "@shape" suffix.

In order to meet that expectation propagating dynamism pass must use output
data object name as prefix. Additionally, propagating pass must be applied
before converting shape notation pass in order to make output shape in IE
notation, not MDK, as MyriadInferRequest::GetResult is expecting.

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>
4 years ago Remove deprecated layer test class (#610)
Mikhail Treskin [Wed, 3 Jun 2020 08:04:15 +0000 (11:04 +0300)]
 Remove deprecated layer test class (#610)

* Update activation layer test

Signed-off-by: Mikhail Treskin <mikhail.treskin@intel.com>
* Get rid of LayerTestsCommonDeprecated class

Signed-off-by: Mikhail Treskin <mikhail.treskin@intel.com>
* Fix activation tests instantiations for gpu and myriad plugins

* Remove leaking inferWithInterp function

4 years ago[IE CLDNN] Activation with fused quantize bug fix (#613)
Mikołaj Życzyński [Wed, 3 Jun 2020 06:30:49 +0000 (08:30 +0200)]
[IE CLDNN] Activation with fused quantize bug fix (#613)

fixed bug connected with quantization fusing to activation
added scale and activation fusing support
added corresponding tests

4 years agoUpdated deprecated messages (#715)
Ilya Lavrenov [Wed, 3 Jun 2020 03:04:50 +0000 (06:04 +0300)]
Updated deprecated messages (#715)

4 years ago[LPT] Output layers fix (#677)
Edward Shogulin [Tue, 2 Jun 2020 20:44:24 +0000 (23:44 +0300)]
[LPT] Output layers fix (#677)

4 years agoEnabled thread tests (#717)
Ilya Churaev [Tue, 2 Jun 2020 20:42:05 +0000 (23:42 +0300)]
Enabled thread tests (#717)

4 years ago[CPU] EmbeddingBagOffsetsSum, EmbeddingBagPackedSum, EmbeddingSegmentsSum operations...
Nikolay Shchegolev [Tue, 2 Jun 2020 18:56:17 +0000 (21:56 +0300)]
[CPU] EmbeddingBagOffsetsSum, EmbeddingBagPackedSum, EmbeddingSegmentsSum operations. (#576)

* [CPU] EmbeddingBagOffsetsSum, EmbeddingBagPackedSum, EmbeddingSegmentsSum operations.

* Performance fix

* Perf v2

* Code style

4 years agoadd PassManagerSettings & create more legible description for concat quantization...
Anna Alberska [Tue, 2 Jun 2020 18:03:27 +0000 (20:03 +0200)]
add PassManagerSettings & create more legible description for concat quantization exception and a test for it (#563)