platform/upstream/dldt.git
3 years ago[CPU] gemm convolution: added post ops JIT implementation (#1817)
Anton Voronov [Wed, 9 Sep 2020 06:17:57 +0000 (09:17 +0300)]
[CPU] gemm convolution: added post ops JIT implementation (#1817)

3 years agoSupport python 3.8 by the Model Optimizer tool in default configuration (#2078)
Roman Kazantsev [Wed, 9 Sep 2020 05:34:43 +0000 (08:34 +0300)]
Support python 3.8 by the Model Optimizer tool in default configuration (#2078)

* Support python 3.8 by the Model Optimizer tool in default configuration

* Fix after review #1

* Fix after the second round review

3 years ago[VPU][OpenCL] Update custom kernels (#2131)
Evgeny Latkin [Wed, 9 Sep 2020 00:50:40 +0000 (03:50 +0300)]
[VPU][OpenCL] Update custom kernels (#2131)

* [Custom CL] Updated OpenCL kernels and tests

* [Custom CL] Update OpenCL compiler

* Update firmware to 1365

* Disable ExpGenerateProposals tests

* VPU: new firmware no. 1370

* Myriad: re-enable ExpGenerateProposals tests

Co-authored-by: Maxim Kurin <maxim.kurin@intel.com>
3 years agoAdd `runPipeline` wrapper to mandatory track full run (#2100)
Vitaliy Urusovskij [Tue, 8 Sep 2020 21:13:07 +0000 (00:13 +0300)]
Add `runPipeline` wrapper to mandatory track full run (#2100)

3 years agoAdded code owners for scripts folder (#2130)
Artyom Anokhov [Tue, 8 Sep 2020 14:23:27 +0000 (17:23 +0300)]
Added code owners for scripts folder (#2130)

3 years agoUpdate tolerance for candy model (#1967)
Mateusz Tabaka [Tue, 8 Sep 2020 14:04:08 +0000 (16:04 +0200)]
Update tolerance for candy model (#1967)

3 years ago[nGraph] Enable u1 data type in Constant PyAPI (#2084)
Jan Iwaszkiewicz [Tue, 8 Sep 2020 11:11:05 +0000 (13:11 +0200)]
[nGraph] Enable u1 data type in Constant PyAPI (#2084)

3 years ago[IE][VPU]: Decompose Swish to Sigmoid + Prod (#2107)
Roman Vyunov (Intel) [Tue, 8 Sep 2020 08:52:05 +0000 (11:52 +0300)]
[IE][VPU]: Decompose Swish to Sigmoid + Prod (#2107)

* Workaround to decompose Swish to Sigmoid + Multiply

3 years ago[IE CLDNN] Switch back to old TI unroller to restore correct matching of TensorIterat...
Mikhail Letavin [Tue, 8 Sep 2020 08:48:11 +0000 (11:48 +0300)]
[IE CLDNN] Switch back to old TI unroller to restore correct matching of TensorIterator output names (#2096)

3 years ago[IE][VPU]: myriad_compile doesn't work without device - fix (#2054)
Daria Mityagina [Tue, 8 Sep 2020 08:21:18 +0000 (11:21 +0300)]
[IE][VPU]: myriad_compile doesn't work without device - fix (#2054)

* Revert setting deprecated PLATFORM config option in compile tool

3 years ago[GNA] add support for NCHW & NHWC layouts for exporting output (#2031)
Anna Alberska [Tue, 8 Sep 2020 07:57:44 +0000 (09:57 +0200)]
[GNA] add support for NCHW & NHWC layouts for exporting output (#2031)

* [GNA] add support for NCHW & NHWC ExportScores

* fix cpplint

3 years ago[GNA] Handling input orientation (#1851)
Andrey Dmitriev [Tue, 8 Sep 2020 07:46:10 +0000 (10:46 +0300)]
[GNA] Handling input orientation (#1851)

Added test

Add fix

3 years agoGRU/RNN/LSTM sequence ops, reference implementations, single layer tests (#1594)
Ivan Tikhonov [Tue, 8 Sep 2020 07:31:44 +0000 (10:31 +0300)]
GRU/RNN/LSTM sequence ops, reference implementations, single layer tests (#1594)

* gru/rnn sequences

* update gru/rnn sequences ops, add unit tests

* enable sequence transformations for cpu plugin

* ngraph codestyle

* update tensor iterator to rnn/gru/lstm sequence transformations, add unit tests

* ngraph codestyle

* add visitors for ngraph ie ops, fix a bug with incorrect axis, fix ngraph to ngraph ie conversion

* update GRUSequence/GRUSequenceIE according to plugin format

* fix ngraph ie implementations according to plugins restricrictions

* fix naming issue

* adapt unit tests to accordance to new changes

* strict checks, additional unit tests

* add descriptions for transformations, fix unit tests

* enable ti to sequnece and unroll transformations in plugins for testing

* disable tensor iterator ngraph reader tests

* delete unnecessary cmake file

* fix includes

* clean up, resolve review comments

* move ti to sequence transformation to ti folder

* validate_and_infer_types() implementation

* input parameter validation for LSTM, GRU and RNN

* style-check applied

* Add LSTMSequence dynamic shape validation and test props for RNNCell, GRUCell, LSTMCell and LSTMSequence.

* recurrent_sequence.hpp moved to ngraph/core/include/ngraph/op/util/

* style check applied

* removed unused variable from LSTMSequence::validate_and_infer_types

* Add missing newline mark at the end of file.

* Add supression macro for FusedOp deprecation.

* Add element type initialization

* transpose,rnn cell reference implementations

* Apply PR review remarks

* reference implementations for cells op, single layer tests, align lstm cell/sequence according to the spec

* lstm/gru/rnn cell decompostion transformations

* ngraph codestyle

* clean up

* ngraph code style

* change inheritance of Cells, fix build

* fix build

* fix build again

* remove Peepholes from LSTMSeq, fix copy_runtime_info in transformations

* Rewrite tests to use gtest exception assertions.

* resolve tests issues

* ngraph codestyle

* add missed files

* fix typeprop tests

* fix lstm sequence checks

* fix arm build

* fix arm again

* delete unnecessary file

* add convert weghts format function, enable lstm test, resolve review comments

* add ngraph builders

* ngraph codestyle

* fix unit tests

* revert transpose reference implementation

* move ti to sequences transformation to another branch, resolve review comments

* resolve review comments

* revert fix in ie_layer_validators

* revert LSTM Cell v0, add LSTMCell v1, update transformation lstm_cell_to_cell_ie

* v1 version of LSTMCell op

* LSTMSequence v1 operation, exclude LSTMSeq from opset4

* fix python api tests

* resolve review comments, tests for decomposition transformations, switch lstm cell to opset4 in mo

* references impl for RNN/GRU/LSTM Sequences, single layer tests, bidirectional transformation

* fix unit tests

* process dynamic ranks of rnn/gru/lstm ops

* remove sequences specifications from opset4

* resolve review comments

* fix validate_and_infer_types of GRU/RNN sequences

Co-authored-by: Szymon Durawa <szymon.durawa@intel.com>
3 years ago[IE CLDNN] Added is_discrete flag into device info and FULL_DEVICE_NAME flag (#2089)
Vladimir Paramuzov [Tue, 8 Sep 2020 06:56:54 +0000 (09:56 +0300)]
[IE CLDNN] Added is_discrete flag into device info and FULL_DEVICE_NAME flag (#2089)

3 years agoONNX Resize fix (#2103)
Tomasz Dołbniak [Tue, 8 Sep 2020 06:51:30 +0000 (08:51 +0200)]
ONNX Resize fix (#2103)

3 years ago[IE][VPU][GT][IE Transformations]: Fixes for post processing model (#2041)
Andrew Bakalin [Mon, 7 Sep 2020 20:23:47 +0000 (23:23 +0300)]
[IE][VPU][GT][IE Transformations]: Fixes for post processing model (#2041)

* Disable StridedSlice to crop convertion in dynamic cases
* NMS: increase shave requirements for some cases
* Update firmware
* Add test cases

3 years ago[LPT] Multiinput with one parent and FQ with three Constant (#2066)
Edward Shogulin [Mon, 7 Sep 2020 17:31:45 +0000 (20:31 +0300)]
[LPT] Multiinput with one parent and FQ with three Constant (#2066)

* [LPT] FakeQuantize with three constants

* [LPT] Dequantization ops on thw inputs with one parent

3 years agoRefactoring: install_openvino_dependencies.sh script (#2112)
Dmitrii Denisov [Mon, 7 Sep 2020 16:54:56 +0000 (19:54 +0300)]
Refactoring: install_openvino_dependencies.sh script (#2112)

3 years ago[IE TESTS] GatherTree single layer test has been created. (#2006)
Maksim Kutakov [Mon, 7 Sep 2020 14:17:14 +0000 (17:17 +0300)]
[IE TESTS] GatherTree single layer test has been created. (#2006)

* [IE TESTS] GatherTree op ref function has been created.

* [IE TESTS] Added GatherTree single layer test

* [IE TESTS] Fixed code styles.

* [IE TESTS] GatherTree test FP32 precion was enabled.

* [IE TESTS] Refactoring of Builder::makeConstatn procedure

The refactoring is aimed at managing the range of random data for the constants initialization procedure.

* [IE TESTS] GatherTree test was extended with constants

* [IE TESTS] GatherTree ref rewritten to non-templated function.

* [IE TESTS] GatherTree test inp shape indx enum removed.

* Revert "[IE TESTS] Refactoring of Builder::makeConstatn procedure"

This reverts commit 2648172e00ccca266d39e8775b890b8a8395f57c.

* [IE TESTS] makeConstant was augmented with random data range parameters.

* [IE TESTS] GatherTree test was rewritten using makeConstant function.

* [IE TESTS] GaterTree test call templated makeConstant

* [IE TESTS] GaterTree test code style fix

3 years agov1::AvgPooling type prop unit tests (#2013)
Tomasz Dołbniak [Mon, 7 Sep 2020 14:05:19 +0000 (16:05 +0200)]
v1::AvgPooling type prop unit tests (#2013)

3 years ago[CPU] fix: supported Tile with more than 5 dims (#2062)
Anton Voronov [Mon, 7 Sep 2020 12:39:19 +0000 (15:39 +0300)]
[CPU] fix: supported Tile with more than 5 dims (#2062)

3 years ago* [CPU] Extend Reduce node to support blocked layouts nC[d]hw8/16C (#580)
Chen Xu [Mon, 7 Sep 2020 12:35:11 +0000 (20:35 +0800)]
* [CPU] Extend Reduce node to support blocked layouts nC[d]hw8/16C (#580)

3 years ago[IE CLDNN] Grouped convolution kernel improvement (#2063)
Mikołaj Życzyński [Mon, 7 Sep 2020 08:52:06 +0000 (10:52 +0200)]
[IE CLDNN] Grouped convolution kernel improvement (#2063)

3 years agoRemove list of the supported ops by ONNX Importer (#2061)
Tomasz Socha [Mon, 7 Sep 2020 08:50:55 +0000 (10:50 +0200)]
Remove list of the supported ops by ONNX Importer (#2061)

3 years ago[IE CLDNN] Fix the order of dimensions in reshapes inserted by Select primitive ...
Mikhail Letavin [Mon, 7 Sep 2020 08:49:56 +0000 (11:49 +0300)]
[IE CLDNN] Fix the order of dimensions in reshapes inserted by Select primitive (#2082)

3 years agoVPU: update firmware version to 1354 (#2087)
Evgeny Latkin [Mon, 7 Sep 2020 07:58:36 +0000 (10:58 +0300)]
VPU: update firmware version to 1354 (#2087)

3 years agoRefactored legacy code for mean_scale_values transformations (#1936)
iliya mironov [Mon, 7 Sep 2020 07:41:47 +0000 (10:41 +0300)]
Refactored legacy code for mean_scale_values transformations (#1936)

* Remove move_to_preproc. Not actual.

* Updated documentation

3 years ago[IE][VPU]: Fix behavior tests for MXpcie (#1879)
Andrey Chekhonin [Mon, 7 Sep 2020 07:38:07 +0000 (10:38 +0300)]
[IE][VPU]: Fix behavior tests for MXpcie (#1879)

* Moved mem type patching under USB protocol switch
* Myriad beh tests fix

3 years agoNgraph improvements (#2058)
Vladislav Volkov [Mon, 7 Sep 2020 07:36:52 +0000 (10:36 +0300)]
Ngraph improvements (#2058)

3 years agoMove FakeOutput resolving to back phase (#2033)
Maxim Vafin [Mon, 7 Sep 2020 07:20:24 +0000 (10:20 +0300)]
Move FakeOutput resolving to back phase (#2033)

3 years ago[IE CLDNN] Fixed fsv16 lrn kernel with fp16 input (#2086)
Vladimir Paramuzov [Mon, 7 Sep 2020 06:04:05 +0000 (09:04 +0300)]
[IE CLDNN] Fixed fsv16 lrn kernel with fp16 input (#2086)

3 years agoHandle Split axes as i64 (#2079)
Tomasz Dołbniak [Mon, 7 Sep 2020 02:48:32 +0000 (04:48 +0200)]
Handle Split axes as i64 (#2079)

3 years ago[IE CLDNN] Fixing blocked format opting for strided_slice (#2073)
Lukasz Debski [Sat, 5 Sep 2020 16:01:43 +0000 (18:01 +0200)]
[IE CLDNN] Fixing blocked format opting for strided_slice (#2073)

3 years ago[IE CLDNN] Disable unsupported blocked formats for softmax (#2049)
Vladimir Paramuzov [Sat, 5 Sep 2020 07:11:18 +0000 (10:11 +0300)]
[IE CLDNN] Disable unsupported blocked formats for softmax (#2049)

3 years ago[IE CLDNN] Reduce performance drop fix (#2065)
Mikołaj Życzyński [Sat, 5 Sep 2020 07:11:00 +0000 (09:11 +0200)]
[IE CLDNN] Reduce performance drop fix (#2065)

3 years ago[IE CLDNN] Enable bfyx fully_connected for large batches (#1996)
Konrad Dobros [Fri, 4 Sep 2020 19:22:46 +0000 (21:22 +0200)]
[IE CLDNN] Enable bfyx fully_connected for large batches (#1996)

This change enables fully_connected to use bfyx format for large
batches, when optimized bf_tiled kernel best performance.

3 years agoFix fusing Multiply node with Convolution in case group != 1 (#1882)
Maxim Vafin [Fri, 4 Sep 2020 17:32:51 +0000 (20:32 +0300)]
Fix fusing Multiply node with Convolution in case group != 1 (#1882)

* Fix fusing Multiply node with Convolution in case group != 1

* Add transformation test

* Do not fuse if not possible to reshape const

* Update fuse_linear_ops.py

3 years agoMake pybind more verbose in debug on windows (#2045)
Tomasz Dołbniak [Fri, 4 Sep 2020 16:03:33 +0000 (18:03 +0200)]
Make pybind more verbose in debug on windows (#2045)

* Make pybind more verbose in debug on windows

* Remove the NDEBUG flag everywhere

* Code complexity reduction...

* Missing colon

* And now the missing empty line...

* Reusable functions

* Now the mood of the sentence was wrong...

* Free functions instead of methods

3 years agoHostTensor2Vector (#2027)
Jozef Daniecki [Fri, 4 Sep 2020 13:02:14 +0000 (15:02 +0200)]
HostTensor2Vector (#2027)

* Add host_tesnor_2_vector() implementation and unit tests. One reference OP refactored to use it.

* Ngraph assertion message refactored.

Co-authored-by: Tomasz Dołbniak <tomasz.dolbniak@intel.com>
* Fix style.

Co-authored-by: Tomasz Dołbniak <tomasz.dolbniak@intel.com>
3 years ago[CPU] Disabled several non-valid static assertions (#2055)
Gorokhov Dmitriy [Fri, 4 Sep 2020 12:38:07 +0000 (15:38 +0300)]
[CPU] Disabled several non-valid static assertions (#2055)

3 years ago[PP GAPI] - Generic precision conversion kernel (#1897)
Anton Potapov [Fri, 4 Sep 2020 12:24:36 +0000 (15:24 +0300)]
[PP GAPI] - Generic precision conversion kernel (#1897)

- U16toF32 conversion kernel converted to more generic ConvDepth one
- U16 <-> F32 conversion only are supported for now
- kernel is not used in the preprocessing graph yet
- tests are extended

3 years agoEnable force shape inference in RNNSeqNormilize transformation (#2069)
Ivan Tikhonov [Fri, 4 Sep 2020 12:08:52 +0000 (15:08 +0300)]
Enable force shape inference in RNNSeqNormilize transformation (#2069)

3 years agofix for typo in src/inference_engine/CMakeLists.txt (#1988)
Polina Brzezinskaya [Fri, 4 Sep 2020 12:07:47 +0000 (15:07 +0300)]
fix for typo in src/inference_engine/CMakeLists.txt (#1988)

3 years ago[IE][VPU]: Introduces annotations of stages memory types (#1991)
Gladilov, Gleb [Fri, 4 Sep 2020 11:33:10 +0000 (14:33 +0300)]
[IE][VPU]: Introduces annotations of stages memory types (#1991)

* Moves splitLargeKernelConv tests to unit tests

Originally, file with tests has been placed in a wrong place
so it was not integrated into any testing application.
Now it is a part of unit tests on VPU.

Test itself has been disabled due to issue with NCE unit usage
described in #-33366

* Introduces pass I/O memory types annotation of stages

It is useful to see where inputs and outputs are located in
performance report for analysing possible issues.

* Introduces endsWith and tuple2Vector utilities

endsWith checks if source has suffix equals to second
argument. tuple2Vector converts tuple of arbitrary size
containing the same type to vector. It could be useful
working with gtest parameter generators that have
std::tuple as return type.

* Introduces unit tests on annotating stages memory types

* Introduces missing format placeholders

* Makes memory types annotation optional

Enables private option "enableMemoryTypesAnnotation" which
disabled by default. Disabling annotation by default allows
avoid issues with tests which rely on stages names.

Signed-off-by: Gladilov, Gleb <gleb.gladilov@intel.com>
3 years ago[GNA] Fix Activation output size not matching convolution if padded. (#1980)
Kamil Magierski [Fri, 4 Sep 2020 10:23:00 +0000 (12:23 +0200)]
[GNA] Fix Activation output size not matching convolution if padded. (#1980)

* Fix Activation output size not matching convolution if padded.

* Fix input padding handling in Convolution

* fix static bug

* Use correct value for feature rotation.

* [GNA] Fix regression

* Added tests

* Added tests

3 years agoFixed speech demo compilation issue on 2021.1 and models are updated to IRv10 (#2029)
Piotr Rozen [Fri, 4 Sep 2020 09:25:45 +0000 (11:25 +0200)]
Fixed speech demo compilation issue on 2021.1 and models are updated to IRv10 (#2029)

3 years ago[IE][VPU]: Reduce XLink log level in AOTBehaviorTests (#2044)
Nikita Kudriavtsev [Fri, 4 Sep 2020 08:59:51 +0000 (11:59 +0300)]
[IE][VPU]: Reduce XLink log level in AOTBehaviorTests (#2044)

3 years ago[IE CLDNN] Fix performance degradation for brain-tumor-segmentation-0002 (#2038)
Vladimir Paramuzov [Fri, 4 Sep 2020 08:54:59 +0000 (11:54 +0300)]
[IE CLDNN] Fix performance degradation for brain-tumor-segmentation-0002 (#2038)

JIRA: 37933

Co-authored-by: Egor Churaev <egor.churaev@intel.com>
3 years ago[IE][VPU]: Fix for Swish layer (#2034)
Nikita Kudriavtsev [Fri, 4 Sep 2020 08:52:13 +0000 (11:52 +0300)]
[IE][VPU]: Fix for Swish layer (#2034)

* Fix Swish layer serialization.
* Added shared test for Swish
* Firmware update 1326 -> 1349

3 years agoAdd Mish with SoftPlus transformation (#1815)
iliya mironov [Fri, 4 Sep 2020 08:07:37 +0000 (11:07 +0300)]
Add Mish with SoftPlus transformation (#1815)

* Add Mish with SoftPlus transformation

* Refactoring accrding code review

* Add softplus to mish pass registration

* Add checks customer count for SoftPlus and Tanh ops

3 years agoHetero plugin dump dot file using ngraph pass (#2042)
Anton Pankratv [Fri, 4 Sep 2020 07:55:40 +0000 (10:55 +0300)]
Hetero plugin dump dot file using ngraph pass (#2042)

* Hetero plugin dump dot file using ngraph pass

* Removed unused code

* Fixed color Index

* reverted prcision forwarding

3 years ago[CPU][BF16] bf16 for Gemm or MatMul was enabled (#1920)
Alexey Varyzgin [Fri, 4 Sep 2020 07:04:02 +0000 (10:04 +0300)]
[CPU][BF16] bf16 for Gemm or MatMul was enabled (#1920)

3 years ago[CPU] Interpolate operation implementation (#1465)
Chenhu Wang [Fri, 4 Sep 2020 06:54:05 +0000 (14:54 +0800)]
[CPU] Interpolate operation implementation (#1465)

3 years ago[IE CLDNN] Fix i8 regression on single-image-super-resolution-1032 (#1990)
Jedrzej Hajduczenia [Fri, 4 Sep 2020 06:20:16 +0000 (08:20 +0200)]
[IE CLDNN] Fix i8 regression on single-image-super-resolution-1032 (#1990)

3 years agoLSTMCell/Sequence v1, reference implementations and decompose transformations for...
Ivan Tikhonov [Fri, 4 Sep 2020 06:04:36 +0000 (09:04 +0300)]
LSTMCell/Sequence v1, reference implementations and decompose transformations for LSTM/GRU/RNN Cells (#2000)

* validate_and_infer_types() implementation

* input parameter validation for LSTM, GRU and RNN

* style-check applied

* Add LSTMSequence dynamic shape validation and test props for RNNCell, GRUCell, LSTMCell and LSTMSequence.

* recurrent_sequence.hpp moved to ngraph/core/include/ngraph/op/util/

* style check applied

* removed unused variable from LSTMSequence::validate_and_infer_types

* Add missing newline mark at the end of file.

* Add supression macro for FusedOp deprecation.

* Add element type initialization

* transpose,rnn cell reference implementations

* Apply PR review remarks

* reference implementations for cells op, single layer tests, align lstm cell/sequence according to the spec

* lstm/gru/rnn cell decompostion transformations

* ngraph codestyle

* clean up

* ngraph code style

* change inheritance of Cells, fix build

* fix build

* fix build again

* remove Peepholes from LSTMSeq, fix copy_runtime_info in transformations

* Rewrite tests to use gtest exception assertions.

* resolve tests issues

* ngraph codestyle

* add missed files

* fix typeprop tests

* fix lstm sequence checks

* fix arm build

* fix arm again

* delete unnecessary file

* add convert weghts format function, enable lstm test, resolve review comments

* add ngraph builders

* ngraph codestyle

* fix unit tests

* revert transpose reference implementation

* revert LSTM Cell v0, add LSTMCell v1, update transformation lstm_cell_to_cell_ie

* v1 version of LSTMCell op

* LSTMSequence v1 operation, exclude LSTMSeq from opset4

* fix python api tests

* resolve review comments, tests for decomposition transformations, switch lstm cell to opset4 in mo

Co-authored-by: Szymon Durawa <szymon.durawa@intel.com>
3 years agoRemoved builders from public API (#2059)
Ilya Churaev [Fri, 4 Sep 2020 03:06:43 +0000 (06:06 +0300)]
Removed builders from public API (#2059)

* Removed builders from public API

* Fixed build

3 years agoFirst time to inference POC (#1964)
Vitaliy Urusovskij [Thu, 3 Sep 2020 19:08:37 +0000 (22:08 +0300)]
First time to inference POC (#1964)

3 years agoReverted conversion of Resize-10 to Interpolate-4 in ONNX Importer. (#2048)
Vladimir Gavrilov [Thu, 3 Sep 2020 16:34:35 +0000 (19:34 +0300)]
Reverted conversion of Resize-10 to Interpolate-4 in ONNX Importer. (#2048)

* Reverted conversion of Resize-10 to Interpolate-4. Now Resize with opset version < 11 generates Interpolate-1 again.

* Corrected tests.

3 years ago[IE][VPU]: allocate maximum shaves for mish activation (#1995)
Evgeny Latkin [Thu, 3 Sep 2020 16:22:56 +0000 (19:22 +0300)]
[IE][VPU]: allocate maximum shaves for mish activation (#1995)

* Mish activation calculation costs more time than memory copy, so
allocate more shaves mish activation.

Co-authored-by: Jiang, Renzhi <renzhi.jiang@intel.com>
3 years ago[IE CLDNN] Fix result storing in leftover's branch (#2050)
Ilya Znamenskiy [Thu, 3 Sep 2020 15:47:32 +0000 (18:47 +0300)]
[IE CLDNN] Fix result storing in leftover's branch (#2050)

3 years ago[IE CLDNN] Fallback to FP16 for non-quantized layers in quantized FP16+INT8 IR (...
Vladimir Paramuzov [Thu, 3 Sep 2020 14:41:29 +0000 (17:41 +0300)]
[IE CLDNN] Fallback to FP16 for non-quantized layers in quantized FP16+INT8 IR (#941)

3 years ago[IE][VPU][Watchdog]: Reduce CPU usage for idle VPU (#2030)
Dmitry Kurtaev [Thu, 3 Sep 2020 13:49:48 +0000 (16:49 +0300)]
[IE][VPU][Watchdog]: Reduce CPU usage for idle VPU (#2030)

3 years agoAVX2/AVX512 32FC1 Resize (#1788)
Anna Khakimova [Thu, 3 Sep 2020 13:22:43 +0000 (16:22 +0300)]
AVX2/AVX512 32FC1 Resize (#1788)

3 years agoFixed visitor for Interpolate-1 and Interpolate-4 (#2051)
Evgeny Lazarev [Thu, 3 Sep 2020 13:00:46 +0000 (16:00 +0300)]
Fixed visitor for Interpolate-1 and Interpolate-4 (#2051)

* Fixed visitor for Interpolate-1 and Interpolate-4

* Code style fix

* Remove unnecessary changes

* Fixed compilation on Linux for Atttribute visitor of vector<size_t>

* Added unit test for IE IR Reader for Interpolate-4

* Updated unit test for IR Reader for Interpolate-4

* Updated unit test

3 years ago[Python Tools] Replace use of layers in cross-check-tools (#1803)
Anastasia Kuporosova [Thu, 3 Sep 2020 12:40:32 +0000 (15:40 +0300)]
[Python Tools] Replace use of layers in cross-check-tools (#1803)

3 years agoFix version print (#2022)
iliya mironov [Thu, 3 Sep 2020 12:39:02 +0000 (15:39 +0300)]
Fix version print (#2022)

* Fix version print

3 years agofixed python & glib vers differences (#2057)
Dmitry S [Thu, 3 Sep 2020 12:10:17 +0000 (15:10 +0300)]
fixed python & glib vers differences (#2057)

3 years agoFixed a case with empty order in transpose op (#2043)
Ilya Churaev [Thu, 3 Sep 2020 11:50:01 +0000 (14:50 +0300)]
Fixed a case with empty order in transpose op (#2043)

* Fixed a case with empty order in transpose op

* Fixed comments

* Fixed transpose tests

* Fixed tests

3 years ago[GNA] Move old headers (#2002)
Pavel Rodionov [Thu, 3 Sep 2020 11:42:30 +0000 (14:42 +0300)]
[GNA] Move old headers (#2002)

3 years ago[GNA] Fixed case of unconnected output of split layer (#1344)
Andrey Dmitriev [Thu, 3 Sep 2020 10:23:58 +0000 (13:23 +0300)]
[GNA] Fixed case of unconnected output of split layer (#1344)

[GNA] Fixed case of unconnected output of split layer

[GNA] Fixed case of unconnected output of split layer

test

[GNA] Fixed case of unconnected output of split layer

fixed

3 years agoFix ConvBiasFusion pass for some rare cases (#2032)
Gleb Kazantaev [Thu, 3 Sep 2020 08:15:50 +0000 (11:15 +0300)]
Fix ConvBiasFusion pass for some rare cases (#2032)

* Extended ConvMultiplyFusion tests with bias fusion

* Updated ConvBiasFusion with precise constant validation

* Fix ConBiasFusion tests

* Updated eltwise fusion tests

* Changed test parameters

3 years agoRemoved attrs prefix for some layers (#2037)
Ilya Churaev [Thu, 3 Sep 2020 04:58:07 +0000 (07:58 +0300)]
Removed attrs prefix for some layers (#2037)

* Removed attrs prefix for some layers

* Fixed tests

3 years ago 38009: InterpolateReshape fix for TensorFlow style Concat (#2047)
Evgenya Stepyreva [Wed, 2 Sep 2020 20:01:12 +0000 (23:01 +0300)]
 38009: InterpolateReshape fix for TensorFlow style Concat (#2047)

* 38009: InterpolateReshape fix fot TensorFlow Models

* Update interpolate_reshape_test.py

* revert debug code

3 years agoMO docs update reshape and deprecate (#2021)
Evgeny Lazarev [Wed, 2 Sep 2020 12:49:32 +0000 (15:49 +0300)]
MO docs update reshape and deprecate (#2021)

* Updated documentation for Reshape, Squeeze and Unsqueeze

* Updated MO documentation about deprecated IR v7

3 years agoAdded Conda CentOS documentation 2020.4 (#2010)
Anton Romanov [Wed, 2 Sep 2020 10:10:14 +0000 (06:10 -0400)]
Added Conda CentOS documentation 2020.4 (#2010)

3 years agoUpdated ConcatOptimization to support Concat with 0D input of one dimension (#2012)
Evgeny Lazarev [Wed, 2 Sep 2020 07:21:23 +0000 (10:21 +0300)]
Updated ConcatOptimization to support Concat with 0D input of one dimension (#2012)

* Updated ConcatOptimization transformation to work when one dimension of input to Concat is 0D

* Fixed ConcatOptimization transformation to reconnect input edges to Concat

* Completely re-written ConcatOptimization

* Updated Concat0D optimization transformation

* Fixed order of traversing Concat input ports

* Refactored ConcatOptimization transformation to use `delete_input_port` function

* Detele trailing unconnected ports in the ConcatOptimization.py

* Cleaner implementation of ConcatOptimization + unit test

3 years ago[IE CLDNN] Fixing eltwise input sizes (#2017)
Lukasz Debski [Wed, 2 Sep 2020 06:36:52 +0000 (08:36 +0200)]
[IE CLDNN] Fixing eltwise input sizes (#2017)

3 years ago[IE CLDNN] Reduce b_fs_yx_fsv16 optimized kernel (#1690)
Mikołaj Życzyński [Wed, 2 Sep 2020 06:35:30 +0000 (08:35 +0200)]
[IE CLDNN] Reduce b_fs_yx_fsv16 optimized kernel (#1690)

3 years ago[IE CLDNN] Add fc kernel for bfyx with batches (#1946)
Konrad Dobros [Wed, 2 Sep 2020 06:32:59 +0000 (08:32 +0200)]
[IE CLDNN] Add fc kernel for bfyx with batches (#1946)

3 years ago[IE CLDNN] Add verbose macros for reorder_inputs (#1947)
Konrad Dobros [Tue, 1 Sep 2020 16:57:33 +0000 (18:57 +0200)]
[IE CLDNN] Add verbose macros for reorder_inputs (#1947)

3 years ago[IE CLDNN] First conv 3d imad (#1935)
Vladimir Paramuzov [Tue, 1 Sep 2020 16:57:13 +0000 (19:57 +0300)]
[IE CLDNN] First conv 3d imad (#1935)

3 years ago[GNA] Improve ReadHeader (#1843)
Pavel Rodionov [Tue, 1 Sep 2020 16:36:56 +0000 (19:36 +0300)]
[GNA] Improve ReadHeader (#1843)

* [GNA] Improve ReadHeader

Issue-34205

* [GNA[ Fix stack corruption issue in ReadHeader

* [GNA] fix code style

Co-authored-by: Kamil Magierski <kamil.magierski@intel.com>
3 years agoFixed replace_output_update_name method to preserve rt info (#1983)
Gleb Kazantaev [Tue, 1 Sep 2020 14:02:46 +0000 (17:02 +0300)]
Fixed replace_output_update_name method to preserve rt info (#1983)

* Fixed replace_output_update_name method to preserve rt info

* added test

3 years agoFixed SoftPlusDecomposition transformation (#2011)
Gleb Kazantaev [Tue, 1 Sep 2020 13:20:42 +0000 (16:20 +0300)]
Fixed SoftPlusDecomposition transformation (#2011)

3 years ago[nGraph] Python tests - move to int32 (#1966)
Jan Iwaszkiewicz [Tue, 1 Sep 2020 13:11:58 +0000 (15:11 +0200)]
[nGraph] Python tests - move to int32 (#1966)

3 years ago[IE CLDNN] ScatterUpdate layer added (#1839)
RomanZm [Tue, 1 Sep 2020 13:07:12 +0000 (16:07 +0300)]
[IE CLDNN] ScatterUpdate layer added (#1839)

3 years ago[GNA] RemovePermutationsNHWCToNCHWPass - check whether permute operation is last...
Anna Alberska [Tue, 1 Sep 2020 12:32:28 +0000 (14:32 +0200)]
[GNA] RemovePermutationsNHWCToNCHWPass - check whether permute operation is last one in the model (#1805)

* [GNA] check whether permute operation is last one in the model

* add assert for checking

* change casting to static

* check casting to ConvolutionLayer

3 years ago[IE CLDNN] Loosen restrictions on channels in 3d fsv16 convolution kernel (#1744)
Lukasz Debski [Tue, 1 Sep 2020 12:17:03 +0000 (14:17 +0200)]
[IE CLDNN] Loosen restrictions on channels in 3d fsv16 convolution kernel (#1744)

3 years ago[IE CLDNN] Disabled choose_impl call for generic nodes in some cases to avoid double...
Vladimir Paramuzov [Tue, 1 Sep 2020 12:00:52 +0000 (15:00 +0300)]
[IE CLDNN] Disabled choose_impl call for generic nodes in some cases to avoid double compilation (#1989)

3 years agoDo not remove convert after the topK (#1950)
Ilya Churaev [Tue, 1 Sep 2020 11:36:30 +0000 (14:36 +0300)]
Do not remove convert after the topK (#1950)

* Do not remove convert after the topK

* Added debug message

* Removed xFail

* Revert "Added debug message"

This reverts commit a01ace4ade88d73e2797b47c58db33943b0f508d.

* Added test

3 years ago[IE][VPU]: Customer model compilation error - fix(#1901)
Roman Vyunov (Intel) [Tue, 1 Sep 2020 10:03:31 +0000 (13:03 +0300)]
[IE][VPU]: Customer model compilation error - fix(#1901)

Fix of 36693 issue.

* Problem: One of the concat inputs is a constant. Adjust_data_layout pass tries to duplicate all inputs that do not meet the strides requirements, and then copy from the original input to the duplicate with strides. But duplicateData with an argument in the form of a constant also creates a constant, and then, when Copy, an error appears, the presence of a constant output, which cannot be.
* Solution: In addConvertedData create an intermediate date with the same description as the constant, and then copy the constant data into it with the required strides.

Co-authored-by: DariaMityagina <daria.mityagina@intel.com>
3 years ago[CPU] CTCLoss operation implementation. (#1482)
Nikolay Shchegolev [Tue, 1 Sep 2020 09:52:31 +0000 (12:52 +0300)]
[CPU] CTCLoss operation implementation. (#1482)

3 years ago[ONNX] Enable ScatterND operator in ONNX importer (#1987)
Katarzyna Mitrus [Tue, 1 Sep 2020 09:10:03 +0000 (11:10 +0200)]
[ONNX] Enable ScatterND operator in ONNX importer (#1987)

* Enable ONNX scatter_nd

* Tests update

3 years agoDynamic shape cell (#1683)
Szymon Durawa [Tue, 1 Sep 2020 09:01:20 +0000 (11:01 +0200)]
Dynamic shape cell (#1683)

* validate_and_infer_types() implementation

* input parameter validation for LSTM, GRU and RNN

* style-check applied

* Add LSTMSequence dynamic shape validation and test props for RNNCell, GRUCell, LSTMCell and LSTMSequence.

* recurrent_sequence.hpp moved to ngraph/core/include/ngraph/op/util/

* style check applied

* removed unused variable from LSTMSequence::validate_and_infer_types

* Add missing newline mark at the end of file.

* Add supression macro for FusedOp deprecation.

* Add element type initialization

* Apply PR review remarks

* Rewrite tests to use gtest exception assertions.

3 years agoIntepolate-4 nGraph operation (#1412)
Vladimir Gavrilov [Tue, 1 Sep 2020 03:57:34 +0000 (06:57 +0300)]
Intepolate-4 nGraph operation (#1412)

* Fixed prototype of evaluate method.

* Rewritten Interpolate-4 ctors (added argument output_shape). Corrected tests.

* Fixed typo.

* Fixed number of args of make_shared in op::v4::Interpolate::clone_with_new_inputs.

* Fixes in Interpolate-4 tests.

* Now ONNX Upsample-1 is readed as Interpolate-4 with 4 inputs.

* Code style fixes.

* Some fixes in Interpolate-4 layer test.

* Now ONNX Upsample-9 is readed as Interpolate-4 with 4 inputs.

* Small fixes.

* Some changes.

* Fixed processing of 'scales' input in evaluation of Interpolate-4: now 'scales' contains scales only from 'axes'.

* Fixes in documentation.

* Now reference implementation of Interpolate-4 is rewritten for using 3 required inputs.

* Some code style fixes.

* Small fix.

* Started to write tests for method evaluate() of Interpolate-4.

* Continued to write tests for evaluate() of Interpolate-4.

* Some fixes.

* Some additions.

* Written draft of tests for 'cubic' mode with using scales.

* Some changes.

* Started to write tests for 'nearest' mode.

* Started to write tests for 'linear_onnx' mode.

* Some changes.

* Small fixes.

* Added setup of output type.

* Small addition.

* Added debug print into Interpolate-4 evaluate.

* Now in Interpolate-4 evaluation tests 3 inputs of Interpolate are Constants.

* Small changes.

* Added some debug print.

* Added more debug print.

* Some fixes.

* Now 4th argument of runtime::interpolate has type std::vector<int64_t>.

* Added checks for result of calculations.

* Added another expected values vector for the mode 'cubic'.

* Temporarily commented result value checks  for the cubic mode.

* Written tests for 'nearest' mode.

* Some reorganization.

* Written tests for 'linear_onnx' mode.

* Fixed padding loop.

* Fixed docs for 'linear_onnx' mode.

* Written tests for 'cubic' mode.

* Deleted debug print.

* Fixed code style.

* Enabled CPU layer tests for Interpolate-4.

* Reverted changes of this file.

* Now ONNX importer reads Resize-1 as Interpolate-4 with 4 inputs.

* Now ONNX importer reads Resize-11 as Interpolate-4 with 4 inputs.

* Small fixes.

* Fixed docs.

* Added small epsilon in the shape calculation in the function op::v4::Interpolate::infer_using_scales.

* Small fix.

* Reduced size of the template function eval().

* Now the 'nearest' mode is rewritten to CoordinateTransfom instead of NDim* classes

* Now the 'cubic' mode is rewritten to use CoordinateTransformation instead of NDim classes.

* Started to rewrite 'linear' mode using CoordinateTransform.

* Started to write helper class.

* Small fix.

* Small changes.

* Some fix.

* Fixed typo.

* Now the preamble of 'linear_onnx' mode implementation is a method of helper class.

* Now the function clip_coord is the method of helper class, and the mode 'linear' uses CoordinateTransform instead of NDim classes.

* Deleted NDim classes.

* Some fixes.

* Some refactoring.

* Some refactoring: now inner calculation of 'linear' is in helper class.

* Moved reference implementation of Interpolate-4 into library with reference implementations.

* Small fix.

* Deleted commented tests.

* Code stile fixes.

* Deleted redundant type prop tests for Interpolate-4.

* Documentation fixes.

* Disabled IE_CPU tests for ONNX Resize: Interpolate-4 is not implemented in plugins.

* Temporarily disabled some ONNX tests.

* Some refactoring: deleted redundant attributes of InterpolateEval class.

* Small fix.

* Added NGRAPH_RTTI_DECLARATION and NGRAPH_RTTI_DEFINITION.

* Added debug print to 'cubic' mode calculation.

* Some deletions.

* Small fix.

* Fixed typos.

* Added another debug print.

* Now indices_shape is constructed from std::vector<std::size_t>(num_of_axes, 4) again.

* Fixed CMakeLists.txt.

* Small fix.

* Added more debug print.

* Fixed typo.

* Fixed calculation of the first argument of helper.clip_coord in the inner loop of cubic_func.

* Some code style fixes.

* Alphabetically sorted operations of opset4.

* Deleted constant cannot_define_axes.

* Used std::iota instead of loop.

* Renamed structure InfoToCallReference.

* Now void op::v4::Interpolate::validate_and_infer_types() checks an element type of an input tensor.

* Code style fix.

* Changes in reading of ONNX Resize and Upsample: we generate Interpolate-4 without 'axes' input.

* Now bodies of functions evaluate_interpolate_v4() and inline bool evaluate() are moved in the method bool op::v4::Interpolate::evaluate.

* Fixes in example of the documentation of Interpolate-4.

* Fixed typos.

* Small fix.

* Some fixes.

* Deleted some type aliases.

* Uncommented a test for 'cubic' mode.

* Small code style fix in bool op::v4::Interpolate::evaluate.

* Uncommented more test for Interpolate-4 reference implementation.

* Added more debug print.

* Some changes.

* Uncommented all tests for Interpolate-4 evaluate().

* Deleted debug print.

* Deleted 'simple' mode from the map nearest_mode_map.

* Code style fixes.

* Disabled some CPU tests.

* Some fixes.

* Small fixes.

* Some fixes.

* Fixed typo.

* Fixed typo.

* Small change.

* Fixed some typos.

* Fixed some typos.

* Fix in operator() of the class GetOriginalCoordinate.

* Disabled some CPU tests.

* Small changes.

* Deleted template function from resize.cpp.

* Code style fixes.

* Refactored the method op::v4::Interpolate::evaluate.

* Added documentation for infer_using_scales() and infer_using_shapes().

* Added documentation for classes GetNearestPixel and GetOriginalCoordinate.

* Small fixes.

* Some code style fixes.

* Small fixes.

* Some changes.

* Added NGRAPH_SUPPRESS_DEPRECATED_START and NGRAPH_SUPPRESS_DEPRECATED_END for using v0::InterpolateAttrs; and using v0::Interpolate;

* Code style fix.

* Enabled ONNX import only tests for Resize-10, Upsample-8, Upsample-9.

* Fixed element type for scales_const. Fixed targetShapes and pads in single layer tests.

* Small changes.

* Added conversion from NGRAPH to CNNLayer.

* Added documentation for the class InterpolateEval.

* Now 'nearest' mode has special tests in cpu single layer tests.

* Small changes.

* Fixes in cpu single layer tests.

* Temporarily commented Interpolate-4 in ConvertFunctionToCNNNetworkTests.

* Added some docs.

* Enabled some tests for Resize-11.

* Added test.

* Corrected expected values in the resize_downsample_scales_align_corners case.

* Added more test for the 'cubic' mode.

* Added more test for linear_onnx mode.

* Deleted debug print for linear_onnx mode.

* Deleted debug prints. Added yet another test for 'nearest' mode.

* Fixes for import of ONNX Resize-10 and Upsamples.

* Applyed Evgeny Lazarev fix for Interpolate-4 infer function.

* Code style fixes.

* Some tests were deleted from unit_test.manifest file for INTERPRETER.

* Deleted test for downscale Resize-10: results of infer are correct, but comparison is not.

* Enabled test INTERPRETER.onnx_empty_initializers_handling.

* Some fixes.

* Added the method run_with_tolerance_as_fp() to the class TestCase and the method compare_results_with_tolerance_as_fp() to the class TestCaseEngine.

* Small fix.

* Small fix.

* Added yet another type_prop test.

* Disabled CPU test IE_CPU.onnx_empty_initializers_handling.

* Code style fixes.

* Enabled some ONNX tests.

* Some changes.

* Code style fix.

* Enabled test INTERPRETER.onnx_model_round.

* Disabled tests with behavior as behavior of INTERPRETER.onnx_resize11_scales_down_linear.

* Changed tolerance for test onnx_empty_initializers_handling.

* Changed tolerance in the test onnx_resize11_sizes_linear (otherwise this test is failed in MacOS). Disabled test INTERPRETER.onnx_resize11_sizes_nearest_asymmetric_floor, because this test failed in MacOS only.

3 years agoPrint friendly name instead of name to stream (#1985)
Ilya Churaev [Tue, 1 Sep 2020 03:04:54 +0000 (06:04 +0300)]
Print friendly name instead of name to stream (#1985)

3 years agoAdded documentation about supported ONNX operations (#2004)
Ilya Churaev [Tue, 1 Sep 2020 03:04:37 +0000 (06:04 +0300)]
Added documentation about supported ONNX operations (#2004)

* Added documentation about supported ONNX operations

* Fixed comments

3 years agoRemoved NGRAPH_IE_ENABLE flag because it is always ON if unit tests are enabled ...
Ilya Churaev [Tue, 1 Sep 2020 03:03:59 +0000 (06:03 +0300)]
Removed NGRAPH_IE_ENABLE flag because it is always ON if unit tests are enabled (#2003)

3 years agoDynamic shape support for builder collapse (#1971)
Gabriele Galiero Casay [Mon, 31 Aug 2020 16:26:36 +0000 (18:26 +0200)]
Dynamic shape support for builder collapse (#1971)

3 years agoRemoved useless export from readers (#2009)
Ilya Lavrenov [Mon, 31 Aug 2020 16:00:58 +0000 (19:00 +0300)]
Removed useless export from readers (#2009)