platform/upstream/dldt.git
3 years agoDisable test ExecGraphSerializationTest.ExecutionGraph_CPU (#3097)
Alexander Perepelkin [Thu, 12 Nov 2020 23:26:55 +0000 (02:26 +0300)]
Disable test ExecGraphSerializationTest.ExecutionGraph_CPU (#3097)

* Disable test ExecGraphSerializationTest.ExecutionGraph_CPU as temporary measure to avoid segfault

* Disable segfaulting getAvailableDevices

3 years agoinstall_NEO_OCL_driver: Added user to render group. (#3099)
Artyom Anokhov [Thu, 12 Nov 2020 16:35:19 +0000 (19:35 +0300)]
install_NEO_OCL_driver: Added user to render group. (#3099)

3 years ago[IE][VPU]: GatherND DTS transformation (#3025)
Andrew Bakalin [Thu, 12 Nov 2020 14:30:41 +0000 (17:30 +0300)]
[IE][VPU]: GatherND DTS transformation (#3025)

* Implement GatherND DTS
* Introduce tests on DTS
* Introduce tests on DSR+GatherND

3 years ago[GNA] Added tests for negative memory layer offset and split - trivial permute -...
Elizaveta Lobanova [Thu, 12 Nov 2020 14:07:38 +0000 (17:07 +0300)]
[GNA] Added tests for negative memory layer offset and split - trivial permute - concat scenario (#2989)

3 years agoadd getting nms_threshold / iou_threshold from RetinaNet (#3075)
Pavel Esir [Thu, 12 Nov 2020 12:04:07 +0000 (15:04 +0300)]
add getting nms_threshold / iou_threshold from RetinaNet (#3075)

* added getting nms_threshold/iou_threshold from original TF RetinaNet model

* iou_threshold definition added

* fixed getting iou_threshold for TF NMS V2, some minor corrections

* added box_encoding to NMS extractors

3 years agoRemoved const_cast from addOutput method (#3091)
Ilya Churaev [Thu, 12 Nov 2020 11:59:28 +0000 (14:59 +0300)]
Removed const_cast from addOutput method (#3091)

3 years ago[CPU] Converted Pad Layer to MKLDNNNode and added more optimized implementations...
Alexandra Sidorova [Thu, 12 Nov 2020 11:25:42 +0000 (14:25 +0300)]
[CPU] Converted Pad Layer to MKLDNNNode and added more optimized implementations (#2586)

Pad-1: fixed ngraph reference for symmetric mode and added cpu tests

3 years agoITT performance counters for first inference (#1741)
Vladislav Volkov [Thu, 12 Nov 2020 11:00:14 +0000 (14:00 +0300)]
ITT performance counters for first inference (#1741)

* ITT performance counters for first inference

ITT counters for nGraph passes

* RTTI for trnasformation passes

* The MKLDNN plugin first inference counters improvements

3 years ago[IE][VPU][TESTS]: support different modes for Interpolate (#2963)
Daria Mityagina [Thu, 12 Nov 2020 10:51:48 +0000 (13:51 +0300)]
[IE][VPU][TESTS]: support different modes for Interpolate (#2963)

3 years agoFixed KW for Windows (#3089)
Ilya Churaev [Thu, 12 Nov 2020 10:34:28 +0000 (13:34 +0300)]
Fixed KW for Windows (#3089)

* Fixed KW for Windows

* Fixed code style

* Fixed compilation

3 years agoMove QueryState from ExecutableNetwork to InferRequest (#2818)
Svetlana Dolinina [Thu, 12 Nov 2020 09:40:43 +0000 (12:40 +0300)]
Move QueryState from ExecutableNetwork to InferRequest (#2818)

* QueryState moved to InferRequest

* deprecate ExecutableNetwork::QueryState,chaged tests (without any check yet)

* fix build

* review fixes + build fix

* build fix + review changes

* remove blank line

* style fixes

* test build fixes

* style fix

* style fix

* fixed build of tests

* fix

* mac build fix

* hddl plugin build fix

* clean up unneeded implementation for method

* fixed tests build

* add implementation for getstate, correct getName for MklDNN

* fixed description of state API in comments

* lint fixes

* Rename MemoryState to VariableState

* added tests for cpu for VariableStates, several small fixes in tests and code

* merge fix

* lint fix

* remove whitespaces

* spaces fix

* fix in test to make it workable for all plugins

* fix typo

* fix test for gna

* remove extra comment

* fix test for gna

3 years agoSummary report generator for layer and subgraph tests (#2707)
Mikhail Treskin [Thu, 12 Nov 2020 09:33:23 +0000 (12:33 +0300)]
Summary report generator for layer and subgraph tests (#2707)

* Initial summary dumper implementation

* Handle Tensoriterator body + add parser script

* Add support of XML reports merging + report OP names with versions

* Remove debug device name change

* Fix windows building issue

* Add --disable_test_skips command line option

* Gtest failure with logging

* Change skipping logic and resolve linkage errors caused by extern

* Get graph body from Loop

* Fix disable_tests_skipping symbol redefinition

* Fix inline for currentTestIsDisabled

* Rollback get_body for Loop

* Handle cases with skip in test SetUp

* Report Loop and TI ops along with ops in subgraph body

* Resolve some PR comments

* Dummy commit to kick pre-commit validation

Co-authored-by: Efode, Irina <irina.efode@intel.com>
3 years agoFixed transformations looking for FusedBatchNorm operation to look for FBNV2 and...
Evgeny Lazarev [Thu, 12 Nov 2020 04:33:39 +0000 (07:33 +0300)]
Fixed transformations looking for FusedBatchNorm operation to look for FBNV2 and FBNV3 also (#3078)

* Fixed transformations looking for FusedBatchNorm operation to consider FusedBatchNormV2 and FusedBatchNormV3 also.

* Updated unit test for FusedBatchNormTraining

* Fixed unit test

3 years ago[IE][VPU]: Implement VPU Ngraph EIP->ReorgYolo transformation (#3066)
Maksim Doronin [Wed, 11 Nov 2020 19:04:59 +0000 (22:04 +0300)]
[IE][VPU]: Implement VPU Ngraph EIP->ReorgYolo transformation (#3066)

* We need to convert ExtractImagePatches op to ReorgYolo to restore the working capacity of myriad plugin while compiling Yolo-v2 models.
* It was previously removed in #2687

3 years agoDon't build extra module if it doesn't have cmake files (#3074)
Ilya Lavrenov [Wed, 11 Nov 2020 17:31:44 +0000 (20:31 +0300)]
Don't build extra module if it doesn't have cmake files (#3074)

3 years ago[LPT] Dequantization constant output precision handling extending (#2987)
Edward Shogulin [Wed, 11 Nov 2020 15:50:37 +0000 (18:50 +0300)]
[LPT] Dequantization constant output precision handling extending (#2987)

* [LPT] moveDequantizationAfter fix

* [LPT] ConcatTransformation fix: only intermediate operations are handled

3 years ago[IE][VPU]: Optimize swish layer and remove swish replacement pass (#2993)
Maxim Kurin [Wed, 11 Nov 2020 14:40:37 +0000 (17:40 +0300)]
[IE][VPU]: Optimize swish layer and remove swish replacement pass (#2993)

* Swish layer optimization
* Update VPU firmware 1468

3 years agoretinanet onnx model has Upsample error (#3052)
Vladimir Gavrilov [Wed, 11 Nov 2020 14:31:00 +0000 (17:31 +0300)]
retinanet onnx model has Upsample error (#3052)

* Commit.

* Fixed infer function of ONNXResize11 operation.

* Small fixes.

* Written unit tests for shape calculation for ONNXResize11.

3 years agoRemoval of obsolete constant folding passes (#2902)
Tomasz Dołbniak [Wed, 11 Nov 2020 12:49:40 +0000 (13:49 +0100)]
Removal of obsolete constant folding passes (#2902)

* Redundant op::Max CF removal

* Redundant op::Min CF removal

* Redundant op::Sum & op::Product CF removal

* CF Min and Max using evaluate()

* Arithmetic reduction CF pass removal

* Quantize op CF pass removal

* Convert op CF pass removal

* Logical reduction CF pass removal

* Select op CF pass removal

* OneHot CF pass removal

* Code formatting

* ScatterElements CF pass removal

* Gather CF pass removal

* Disable a Quantize op test that fails in CI

* CF pass cleanup

* Convert op cleanup and test adaptation to spec

* Possible fix for failing VPU tests

* Limit the types used in OneHot::evaluate

* Quantize op evaluator removal

* Refactor of Gather evaluator

3 years agoFix spelling errors in samples and documentation (#2795)
Roman Donchenko [Wed, 11 Nov 2020 12:35:39 +0000 (15:35 +0300)]
Fix spelling errors in samples and documentation (#2795)

* Fix spelling errors in samples

* Fix spelling errors in the documentation

3 years ago[LPT] CNNNetwork LPT removal (#3035)
Edward Shogulin [Wed, 11 Nov 2020 11:56:58 +0000 (14:56 +0300)]
[LPT] CNNNetwork LPT removal (#3035)

* [LPT] CNNNetwork LPT removal

* Fixes for LPT removal

Co-authored-by: Ilya Lavrenov <ilya.lavrenov@intel.com>
3 years agoFix GatherND to output dynamic rank tensor (#3073)
Roman Kazantsev [Wed, 11 Nov 2020 11:47:00 +0000 (14:47 +0300)]
Fix GatherND to output dynamic rank tensor (#3073)

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
3 years ago[IE][VPU][Tests]: Fix NMS DTS outputs naming + tests (#3040)
Andrew Bakalin [Wed, 11 Nov 2020 10:52:42 +0000 (13:52 +0300)]
[IE][VPU][Tests]: Fix NMS DTS outputs naming + tests (#3040)

* Fix dynamic output case in interpreterFunction. For dynamic output cases, we can't call get_shape on the result because it's shape is dynamic, instead, we should take the real output shape from output HostTensor
* Fix outputs naming as it's done in other DTS transformation for operations with multiple outputs (Split, TopK, etc).

Ticket - #-42421

3 years ago[IE CLDNN] Add asymmetric quantization support to reference convolution imad kernel...
Jedrzej Hajduczenia [Wed, 11 Nov 2020 09:38:45 +0000 (10:38 +0100)]
[IE CLDNN] Add asymmetric quantization support to reference convolution imad kernel (#2994)

3 years agoAdded support for a new version of the TF OD API pre-processing part (#3063)
Evgeny Lazarev [Wed, 11 Nov 2020 08:53:10 +0000 (11:53 +0300)]
Added support for a new version of the TF OD API pre-processing part (#3063)

* Added support for a new version of the TF OD API pre-processing part of the mode

* Get rid of legacy API usage

* Fix comment and added assert

* Wording

3 years ago[IE CLDNN] Added LogSoftmax-5 operation (#2945)
Roman Lyamin [Wed, 11 Nov 2020 05:53:30 +0000 (08:53 +0300)]
[IE CLDNN] Added LogSoftmax-5 operation (#2945)

3 years agoFix wrong statement in ConvertConvolutions transformation (#3056)
Gleb Kazantaev [Wed, 11 Nov 2020 04:49:22 +0000 (07:49 +0300)]
Fix wrong statement in ConvertConvolutions transformation (#3056)

* Fixed wrong statement in ConvertConvolution transformation

* Added tests

* FIXME: 42956

3 years ago[CPU] Proposal layer. Check img info input. (#2456)
Nikolay Shchegolev [Tue, 10 Nov 2020 21:36:38 +0000 (00:36 +0300)]
[CPU] Proposal layer. Check img info input. (#2456)

3 years ago[GNA] Added Multiplication 1x1x1 (#2470)
Andrey Dmitriev [Tue, 10 Nov 2020 17:56:21 +0000 (20:56 +0300)]
[GNA] Added Multiplication 1x1x1 (#2470)

* [GNA] Broadcast power primitive

[GNA] Broadcast power primitive

[GNA] Added Multiplication 1x1x1

[GNA] Broadcast power primitive

Added secondary changes

* Added fix test

3 years ago[IE][VPU]: support new operation CEILING (#3004)
Andrey Sokolov [Tue, 10 Nov 2020 16:28:53 +0000 (19:28 +0300)]
[IE][VPU]: support new operation CEILING (#3004)

Add new Operation "Ceiling" for VPU Myriad
task: -42885

3 years agoReshape-able SliceConverter (#2954)
Yegor Kruglov [Tue, 10 Nov 2020 14:51:59 +0000 (17:51 +0300)]
Reshape-able SliceConverter (#2954)

* initial commit

* add cast

* data type fix

* added tests

* added test without axes and steps

* remove redundant imports

* discussions resolving

* Add cast to TFSliceToSlice

* layer tests fix

* update unittest

3 years agoProvide GatherND with original layout for inputs and output (#3002)
Roman Kazantsev [Tue, 10 Nov 2020 14:24:04 +0000 (17:24 +0300)]
Provide GatherND with original layout for inputs and output (#3002)

* Provide GatherND with original layout for inputs and output

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix code review #1

Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
3 years ago[GNA] fixed case when input connected to concat from left and const from right (...
Andrey Dmitriev [Tue, 10 Nov 2020 13:37:03 +0000 (16:37 +0300)]
[GNA] fixed case when input connected to concat from left and const from right (#2874)

* Added test

* Added fix

* Fix comment

* Fix conflict

3 years ago[GNA] Added fix for export scores (#3009)
Andrey Dmitriev [Tue, 10 Nov 2020 13:33:41 +0000 (16:33 +0300)]
[GNA] Added fix for export scores (#3009)

3 years agoOneCore toolchain (#3044)
Ilya Lavrenov [Tue, 10 Nov 2020 13:29:37 +0000 (16:29 +0300)]
OneCore toolchain (#3044)

* Added OneCoreUap toolchain file

* Intentation

* Updated OneCore toolchain

* Fixed linkage error with static runtime

* Link reader_api library

* Enable static runtime for ONNX part

* Allow to override flags

* Added flag for clDNN; plus refactoring

* Fix Includes

* Added toolchain check

* Added apivalidator for MYRIAD plugin

* Added commented apivalidator for CLDNN plugin

3 years agoReference implementations for Loop and TensorIterator ops (#2978)
Ivan Tikhonov [Tue, 10 Nov 2020 12:49:59 +0000 (15:49 +0300)]
Reference implementations for Loop and TensorIterator ops (#2978)

* Loop op ngraph implementation, update IE IR Reader and ngraph to cnn converter

* refactoring SubGraphOp class

* type prop unit tests

* ngraph code style

* update comment

* single layer tests for Loop operation

* fix file name

* Add SpecialBodyPorts attribute in Loop op, update single layer tests

* first debug version

* more tests

* missing test file

* removed not needed shapes from test data

* move test data to new folder

* shape infer tests

* Added execution tests

* add several new tests cases, strict checks in Loop impl, temporary disable single layer tests

* ngraph codestyle, refactoring, clone_new_args test

* resolve review remarks

* fix build

* fix tests

* more execution tests

* add a new constructor of Loop op, resolve review remarks

* execution tests

* synchro with current version

* handle scalars and more tests

* scalar test enabled

* loop reference impl

* bug fixes in tests, onnx importer part and in the ref implementation of the Loop op

* applied remarks

* handle unsupported cases

* rewrite unit tests

* update INTERPRETER manifest

* is_termination_condition_always_true simplification

* [TEST] update python models tests

* review remarks

* added xfail to tiny_yolov3

* missing model test

* revert test data

* fixed numbers of failing tests

* fixed failed test description

* fix test message

* fix xfail test

* reference implementation for ngraph::function

* update loop reference implementation

* Refactor loop reference implementation

* ngraph codestyle

* Refactoring

* Submodule update

* Skip check for Reduce ops in mkl for scalar cases, support for yolov3

* fix ngraph reader tests

* revert ceiling op, renaming

* Add allias(Ceiling) for Ceil op in mkl

* delete xfails

* fix build

* single layer tests for tensor iterarator

* Refactor TensorIterator and Loop ref impls

* revert dynamic tensor creation, disable some dynamic test cases

* fix warning

* Resolve review remarks

* revert Predefined values in Loop tests

Co-authored-by: Mateusz Bencer <mateusz.bencer@intel.com>
3 years agoAdd fix for multiple_activations test (#2984)
Krzysztof Bruniecki [Tue, 10 Nov 2020 10:40:28 +0000 (11:40 +0100)]
Add fix for multiple_activations test (#2984)

* Add fix for multiple_activations test

   - Add forbid activation fusing pass for GNA2 library
   - Fix get all prev layers fnct
   - To enable GNA_NoRegression.smoke_multiple_activations_onGNA_INT16

* Apply Bartek review

3 years ago[IE][Myriad] Fixing global_lock_fd locking in mvnc (#3024)
Nikita Kudriavtsev [Tue, 10 Nov 2020 09:39:32 +0000 (12:39 +0300)]
[IE][Myriad] Fixing global_lock_fd locking in mvnc (#3024)

3 years agoSet friendly name in order to operate with it on visit attributes and validate stages...
Ilya Churaev [Tue, 10 Nov 2020 09:24:53 +0000 (12:24 +0300)]
Set friendly name in order to operate with it on visit attributes and validate stages (#3048)

3 years ago[ONNX Importer] Resolved problem with incorrect opset version set in subgraph (#2959)
Mateusz Bencer [Tue, 10 Nov 2020 08:57:06 +0000 (09:57 +0100)]
[ONNX Importer] Resolved problem with incorrect opset version set in subgraph (#2959)

3 years ago[IE TESTS] Replaced size_t with int64_t (#2699)
Liubov Batanina [Tue, 10 Nov 2020 08:35:24 +0000 (11:35 +0300)]
[IE TESTS] Replaced size_t with int64_t (#2699)

* Replaced size_t with int64_t

* Fix gpu plugin

3 years ago[CPU] fixed strides filling in blocking desc (#2353)
Anton Voronov [Tue, 10 Nov 2020 08:25:12 +0000 (11:25 +0300)]
[CPU] fixed strides filling in blocking desc (#2353)

3 years agoUpdate get started Linux instructions (#3016)
Jesus Espinoza [Tue, 10 Nov 2020 07:57:27 +0000 (23:57 -0800)]
Update get started Linux instructions (#3016)

* Updating broken link on getting started linux doc

Link to build instructions was broken, updated link to the correct location.

* Update get-started-linux.md

* Update get-started-linux.md

* Update get-started-linux.md

* Update get-started-linux.md

Fixed broken links and updated commands

3 years ago[CPU] Fixed Eltwise node endless loop in case nthreads = 1 (#3008)
Gorokhov Dmitriy [Tue, 10 Nov 2020 07:56:22 +0000 (10:56 +0300)]
[CPU] Fixed Eltwise node endless loop in case nthreads = 1 (#3008)

3 years agoRevise Broadcast reference implementation (#2715)
Mateusz Bencer [Tue, 10 Nov 2020 07:42:26 +0000 (08:42 +0100)]
Revise Broadcast reference implementation (#2715)

* change tile reference implementation

* remove tile tests from interpreter manifest

* add repeats parameter to tile

* improve tile reference implementation

* add repeats parameter to tile reference call in tile evaluate method

* style apply

* include <numeric>

* add unnamed namespace to helper functions. Change stdio.h to cstdio. Change input_rank to be constant int

* add const reference to parameter repeats in tile reference function

* change createPitches function to use partial_sum instead of accumulate

* change a little bit createPitches function

* style-apply

* fix function naming

* style-apply

* fix calling functions name bug

* Add description of create_pitches function

* first version with debug logs

* reduce footprint

* single layer tests

* added more tests

* fixed handling bool type

* styles applied

* fix tile

* [ONLY DEBUG] print error scenario message

* fixed problem with e2e tests

* fixed casting of start_axis for numpy mode

Co-authored-by: pszmel <piotr.szmelczynski@intel.com>
3 years ago[ONNX] Return NMS-5 generation (#3003)
Andrew Bakalin [Tue, 10 Nov 2020 06:45:14 +0000 (09:45 +0300)]
[ONNX] Return NMS-5 generation (#3003)

* [ONNX] Return NMS-5 generation

* Disable failed models

3 years agoUpdated MO requirements for numpy (#3034)
Evgeny Lazarev [Tue, 10 Nov 2020 06:12:09 +0000 (09:12 +0300)]
Updated MO requirements for numpy (#3034)

3 years agoApply rest of the comments for NMS-5 review (#3037)
Vladimir Gavrilov [Tue, 10 Nov 2020 04:18:40 +0000 (07:18 +0300)]
Apply rest of the comments for NMS-5 review (#3037)

* Commit.

* Moved the preprocessing functions for NMS-5 to Interpreter backend.

* Code style fix.

3 years agoRemoved legacy dependency in template_extensions (#3026)
Ilya Lavrenov [Mon, 9 Nov 2020 17:48:46 +0000 (20:48 +0300)]
Removed legacy dependency in template_extensions (#3026)

3 years ago[Jenkinsfile] Get rid of dldtPipelineEntrypoint (#3012)
Alina Kladieva [Mon, 9 Nov 2020 16:17:19 +0000 (19:17 +0300)]
[Jenkinsfile] Get rid of dldtPipelineEntrypoint (#3012)

* [Jenkinsfile] Get rid of dldtPipelineEntrypoint

* [Jenkinsfile] Add library_version parameter

Co-authored-by: akladiev <akladiev@nncv-akladieva.inn.intel.com>
3 years ago[GNA] Add support for SubstituteSoftSign for pattern with addition (#2789)
Anna Alberska [Mon, 9 Nov 2020 13:40:31 +0000 (14:40 +0100)]
[GNA] Add support for SubstituteSoftSign for pattern with addition (#2789)

* [GNA] add changes for SubstituteSoftSign for IRv10

* cpplint fix

* [GNA] add requested changes & change SoftSign test validation

* [GNA] SubstituteSoftsign refactor

* [GNA] enable 3d input tests for softsign

3 years agofix for hetero example (https://jira.devtools.intel.com/browse/CVS-33836) NB: the...
Maxim Shevtsov [Mon, 9 Nov 2020 11:42:18 +0000 (14:42 +0300)]
fix for hetero example (https://jira.devtools.intel.com/browse/CVS-33836) NB: the FPGA (even though deprecated) continues to be the only viable example foir the HETERO plugin, so leaving that (#2982)

3 years agoFix versions (#3027)
Rafal Blaczkowski [Mon, 9 Nov 2020 09:55:15 +0000 (10:55 +0100)]
Fix versions (#3027)

Have some concerns on numpy version, probably should be revisited further

3 years ago[CPU] Refactored pushInput (#2910)
Maxim Andronov [Mon, 9 Nov 2020 05:52:54 +0000 (08:52 +0300)]
[CPU] Refactored pushInput (#2910)

3 years agoRemove legacy files (licenses and some scripts) (#2952)
Ilya Churaev [Mon, 9 Nov 2020 03:39:52 +0000 (06:39 +0300)]
Remove legacy files (licenses and some scripts) (#2952)

* Remove legacy files (licenses and some scripts)

* Removed LISENSE and install rules for licenses

3 years agoExtend MO for the operation NonMaxSuppression-5 (#2356)
Vladimir Gavrilov [Fri, 6 Nov 2020 18:00:55 +0000 (21:00 +0300)]
Extend MO for the operation NonMaxSuppression-5 (#2356)

* Commit.

* Written draft of NonMaxSuppression-5 class.

* Written conversion of the value of the second output  of MO NonMaxSuppression-5 into TF format.

* Fixed type infer for the port 1 of NonMaxSuppression-5.

* Added Reshape to [1] for 0D inputs of NMS-5.

* Small fix.

* Corrected assert for number of inputs.

* Fixed docstrings for transformations TFNonMaxSuppressionNormalize and NonMaxSuppressionNormalize.

* Now the transformation TFNonMaxSuppressionNormalize uses find_and_replace_pattern().

* Moved model-optimizer/extensions/front/onnx/non_max_suppression_normalize.py to model-optimizer/extensions/front/non_max_suppression_normalize.py, to delete duplicate code.

* Deleted commented code.

* Fixed BOM-file.

* Deleted out_ports_count from NMS.

* Fixes in type_infer of NMS-5.

* Small changes.

* Added some comment.

* Small fix.

* Some fixes.

3 years agoEnable NonMaxSuppression-5 operation and various transformations (#2450)
Vladimir Gavrilov [Fri, 6 Nov 2020 16:11:04 +0000 (19:11 +0300)]
Enable NonMaxSuppression-5 operation and various transformations  (#2450)

* Some code style fixes.

* Started to write the method v5::NonMaxSuppression::evaluate().

* Started to write nGraph reference implementation of NMS-5.

* Some additions.

* Written preprocessing of boxes data.

* Started to write the function non_max_suppression() that calculates NonMaxSuppression-5 operation.

* Written postprocessing of the evaluate().

* Small fixes.

* Small fix.

* Added include for ngraph/shape.hpp.

* Written the function intersectionOverUnion.

* Some additions.

* Small fix.

* Continued to write the function non_max_suppression().

* Small fixes.

* Small fixes.

* Some changes.

* Some additions.

* Some replacements size_t by int64_t.

* Added casts to float in the construction of selected_score variable.

* Code style fixes.

* Written draft of NMS-5 nGraph reference implementation.

* Code style fixes.

* Started to write tests for void op::v5::NonMaxSuppression::validate_and_infer_types().

* Added tests for scalars/nonscalars.

* Fixes in the test type_prop.nms_v5_output_shape.

* Fixes in tests nms_v5_output_shape_2 and nms_v5_output_shape.

* Written tests for validate_and_infer_types() of NMS-5.

* Code style fixes.

* Now NMS-5 evaluate() can have outputs with calculated shapes.

* Small fixes.

* Corrected tests for NMS-5 validate_and_infer_type().

* Code style fixes.

* Started to write inner version of NMS-5 with static output shapes.

* Written draft of the inner operation NonMaxSuppressionIE3.

* Started to write conversion of op::v5::NonMaxSuppression into NonMaxSuppressionIE3.

* Small changes.

* Some additions.

* Small fixes.

* Fixed typo.

* Fixed typos.

* Written draft of the transformation ConvertNMS5ToLegacyMatcher that converts ngraph::opset5::NonMaxSuppression into op::NonMaxSuppressionIE3.

* Written header file for the transformations from NMS-1, NMS-3, NMS-4 to NMS-5.

* Started to write conversion of NMS-4 to NMS-5.

* Added include for ngraph/opsets/opset4.hpp.

* Started to write conversion of NMS-3 to NMS-5.

* Small fixes.

* Written draft of the conversion of NMS-3 into NMS-5.

* Fixed typo.

* Started to write conversion of NMS-1 to NMS-5.

* Written draft of the conversion NMS-1 to NMS-5.

* Started to write tests for the conversion nGraph NMS-5 to inner NMS.

* Started to write the test ConvertNMS5ToNMSIEStatic.

* Written tests for conversion of nGraph NMS-5 to inner NMSIE3.

* Started to write tests for conversion of previous NMS to nGraph NMS-5.

* Written tests for conversion of old nGraph NMS to NMS-5.

* Started to write tests for opset5::NonMaxSuppression::evaluate().

* Some additions.

* Small fix.

* Written tests for op::v5::NonMaxSuppression::evaluate().

* Used NGRAPH_RTTI_DECLARATION for NonMaxSuppressionIE3.

* Used NGRAPH_RTTI_DECLARATION for NMS-5.

* All static local constants and functions for NMS-5 were moved into non-named namespace.

* Some code style fixes.

* Moved some file.

* Small fix.

* Code style fix.

* Now NMS-5 supports all floating types in inputs 0 and 1.

* Moved some files.

* Fixed include directive in the file convert_nms_5_to_legacy.cpp with transformations NMS-1, 3, 4 -> NMS-5.

* Small changes.

* Deleted conversion NMS-3 -> legacy.

* Small changes.

* Fix in op::v5::NonMaxSuppression::evaluate: output shape [1] instead of [] in the output port 2.

* Code style fixes.

* Deleted conversion of NMS-4 into legacy NMS.

* Deleted redundant ifs.

* Added NMS-5 to Python API.

* Code style fix.

* Small change.

* Fixed element type for constants in the conversion of NMS-5 to NMSIE3.

* Deleted support of f64 in NMS-5.

* Added checks for input element types for inputs #0, #1, #3, #4, #5.

* Small change.

* Now get_floats throws an exception for unsupported types.

* Now nGraph NMS-5 supports 0D and 1D tensors in inputs #2, #3, #4, #5.

* Small fix in test_non_max_suppression.

* Deleted using namespace std

* Fixes in test_non_max_suppression().

* Small fixes.

* Added 'import PartialShape' in test_reduction.py.

* Deleted creating fake inputs in the ctor of op::v5::NonMaxSuppression.

* Deleted creating fake inputs in op::v5::NonMaxSuppression::clone_with_new_inputs.

* Corrections in int64_t op::v5::NonMaxSuppression::max_boxes_output_from_input() const.

* Corrected functions float op::v5::NonMaxSuppression::soft_nms_sigma_from_input() const, float op::v5::NonMaxSuppression::score_threshold_from_input() const,  float op::v5::NonMaxSuppression::iou_threshold_from_input() const.

* Small fixes.

* Deleted commented code.

* Fixes in nms_v5_scalar_inputs_check.

* Some changes.

* Small fixes.

* Code style fixes.

* Small changes.

* Small changes.

* Small fix.

* Deleted commented code.

* Some refactoring in ConvertNMS4ToNMS5 ctor.

* Small fix.

* Common part of conversions NMS-1 -> NMS-5, NMS-3 -> NMS-5, NMS-4 -> NMS-5 was moved into the separate function.

* Now conversions NMS-1 -> NMS-5, NMS-3 -> NMS-5, NMS-4 -> NMS-5 support both 2 inputs, and 5 inputs.

* Now transformations NMS-1 -> NMS-5, NMS-3 -> NMS-5, NMS-4 -> NMS-5 are called from 'umbrella' transformation.

* Now the transformation ConvertNMS5ToLegacyMatcher supports NMS-5 with 2, 3, 4, 5, or 6 inputs.

* The transformation ConvertNMS5ToLegacyMatcher was rewritten using Reshape instead of Unsqueeze.

* Started to rewrite tests for the transformation ConvertNMS5ToLegacyMatcher.

* Some fixes.

* Small fixes.

* Corrected tests for the transformation NMS-5 -> NMSIE3.

* Small formatting fix.

* Now methods max_boxes_output_from_input(), iou_threshold_from_input(), score_threshold_from_input(), soft_nms_sigma_from_input() of op::v5::NunMaxSuppression are public.

* Started to move op::v5::NonMaxSuppression::evaluate() into ngraph/test/runtime/interpreter.

* Added NMS-5 to ngraph/test/runtime/interpreter/int_executable.

* Small fixes.

* Code style fixes.

* Written draft test nonmaxsuppression_center_point_box_format_backend in ngraph/test/backend.

* Small fix.

* Written draft tests of NonMaxSuppression in ngraph/test/runtime.

* Some changes.

* Small changes.

* Disabled IE_CPU tests for NMS-5.

* Deleted op_eval tests for NMS-5.

* Deleted evaluate() method of NMS-5.

* Now all nGraph functions in tests of the transformation NMS-5 -> NMSIE3 have one output.

* Now preprocessing and postprocessing of the calculation of NMS-5 in reference implementation.

* Code style fixes.

* Some fixes in tests for the transformation NMS-5 -> NMSIE3.

* Replaced precision i64 -> i32 for some constants in tests for the transformation NMS-5 -> NMSIE3.

* Written creating CNNLayer for NMS-5.

* Added creating CNNLayer for NonMaxSuppressionIE3.

* some changes.

* Now conversions NMS-1, NMS-3, NMS-4 -> NMS-5  and NMS-5 -> NMSIE3 generate NMS nodes with 5 inputs.

* Fixed ctor in MKLDNN NonMaxSuppressionImpl: validation of number of output edges.

* Added conversion of output_type for NMS-5.

* Fixes in the transformation NMS5 -> NMSIE3.

* Fixes in the conversion of NMS-5 to NMSIE3.

* Fixes in MKLDNN NMS ctor.

* Small fix.

* Fixed tests for the transformation NMS5 -> NMSIE3.

* Fixed tests for conversions NMS-1, NMS-3, NMS-4 -> NMS-5.

* Small fixes in MKLDNN NMS ctor.

* Rewritten create_layer() functions for NMS-5 and NMSIE3 as addSpecificCreator() functions.

* Disabled tests for IE IR reader for NMS-4.

* Deleted debug code.

* Added comment about disabling tests IE_CPU.nonmaxsuppression.

* Written IE IR Reader test for NMS-4.

* Deleted function float_from_constant_node.

* Small fixes.

* Deleted functions function_from_model and construct_weights.

* Small fix.

* Replaced push_back with emplace_back in the conversion of NMS-5 to legacy.

* Small changes.

* Some fixes.

* Refactored reference implementation of NMS-5.

* Moved structure NMSAttributes to unnamed namespace.

* Code style fixes.

* Small fix.

3 years agoFix the NHWC->NCHW transformation for dynamic weights (#2848)
Maxim Vafin [Fri, 6 Nov 2020 16:04:46 +0000 (19:04 +0300)]
Fix the NHWC->NCHW transformation for dynamic weights (#2848)

* Fix the NHWC->NCHW transformation when weights and data comes from same input

* Simplify code

3 years agoMatMul spec update (#2765)
Katarzyna Mitrus [Fri, 6 Nov 2020 15:41:36 +0000 (16:41 +0100)]
MatMul spec update (#2765)

Aligned specification with nGraph MatMul shape inference operation. Affects description of the behavior for 1D tensors.

3 years ago[LPT] integration branch: Reshape fix, Concat generalization, runtime info usage...
Edward Shogulin [Fri, 6 Nov 2020 13:15:27 +0000 (16:15 +0300)]
[LPT] integration branch: Reshape fix, Concat generalization, runtime info usage extending (#2930)

* [LPT] Concat transformation generalization

* [LPT] Reshape transformation fix

* [LPT] Legacy callback fix

* [LPT] * added rt_info propagation
      * functional tests: added rt_info
      * functional tests: added MoveDequatnizationAfter tests

Co-authored-by: Vladislav Golubev <vladislav.golubev@intel.com>
3 years agoLow Latency transformation (#2869)
Ivan Tikhonov [Fri, 6 Nov 2020 11:11:11 +0000 (14:11 +0300)]
Low Latency transformation (#2869)

* initial draft of adding sinks to ngraph::Function

* style fixes

* code style fixes

* code style fixes

* code style fix

* review fix+build fix

* code style fix

* fix build

* API changed according to latest discussion

* review fixes

* review fixes + tests

* initial draft of adding sinks to ngraph::Function

* style fixes

* code style fixes

* code style fixes

* code style fix

* review fix+build fix

* code style fix

* fix build

* API changed according to latest discussion

* review fixes

* review fixes + tests

* added 1 more ctor

* style fixes

* used new api in ir parser

* fixed build

* update low latency transformation, fix unroll transformation, add unit tests, modify subgraph tests

* fix low latency transformation

* Update low latency transformation, unit and sub-graph tests

* update LowLatency transformation and tests

* ngraph codestyle

* fix build, update description

* resolve review remarks

Co-authored-by: Svetlana Dolinina <svetlana.a.dolinina@intel.com>
3 years ago[OpenCL] movi-cltools-20.09.2 (#2912)
Maxim Kurin [Fri, 6 Nov 2020 10:45:45 +0000 (13:45 +0300)]
[OpenCL] movi-cltools-20.09.2 (#2912)

3 years ago[CPU] Added LogSoftmax-5 support (#2743)
Alexandra Sidorova [Fri, 6 Nov 2020 10:22:02 +0000 (13:22 +0300)]
[CPU] Added LogSoftmax-5 support (#2743)

3 years agoRemove deprecated v0::TopK (#2947)
Mateusz Tabaka [Fri, 6 Nov 2020 08:36:33 +0000 (09:36 +0100)]
Remove deprecated v0::TopK (#2947)

* Remove deprecated v0::TopK

* Remove using SortType and using Mode from TopK

* apply code format

* Revert "Remove using SortType and using Mode from TopK"

This reverts commit ee2f99ea89583f0cc5c34fe633dfbe900f4d1741.

3 years agoRemove deprecated v0::Softmax (#2917)
Mateusz Tabaka [Fri, 6 Nov 2020 08:34:23 +0000 (09:34 +0100)]
Remove deprecated v0::Softmax (#2917)

3 years agoFeature/azaytsev/layout updates (#2985)
Andrey Zaytsev [Fri, 6 Nov 2020 08:07:53 +0000 (11:07 +0300)]
Feature/azaytsev/layout updates (#2985)

* Reconfigured the layouts with importing the dedicated POT layout and Accuracy Checker and Model Downloader items from the OMZ layout

* updated

* test

* test

* Moved the Model Downloader from ie_docs.xml to openvino_docs.xml

* Renamed optimization_docs.xml --> pot_docs.xml

* Added Model Server to the layout

* Minor fixes

3 years ago[IE CLDNN] Removed unused conv kernels (#2995)
Vladimir Paramuzov [Fri, 6 Nov 2020 07:24:48 +0000 (10:24 +0300)]
[IE CLDNN] Removed unused conv kernels (#2995)

3 years ago[CPU] Added support NMS-5 (#2768)
Maxim Andronov [Fri, 6 Nov 2020 05:59:23 +0000 (08:59 +0300)]
[CPU] Added support NMS-5 (#2768)

3 years ago[IE CLDNN] Plugin-side kernels caching (#2871)
Vladimir Paramuzov [Thu, 5 Nov 2020 20:44:45 +0000 (23:44 +0300)]
[IE CLDNN] Plugin-side kernels caching (#2871)

3 years agoovms doc fix (#2988)
Trawinski, Dariusz [Thu, 5 Nov 2020 18:19:54 +0000 (19:19 +0100)]
ovms doc fix (#2988)

* added OpenVINO Model Server

* ovms doc fixes

3 years agoAdded support for the Loop operation in the MO IR Reader (#2971)
Evgeny Lazarev [Thu, 5 Nov 2020 17:53:44 +0000 (20:53 +0300)]
Added support for the Loop operation in the MO IR Reader (#2971)

* Added support for the Loop operation in the MO IR Reader

3 years ago[GNA] support for 3D [1,1,X] import/output (#2893)
Anna Alberska [Thu, 5 Nov 2020 16:31:55 +0000 (17:31 +0100)]
[GNA] support for 3D [1,1,X] import/output (#2893)

* [GNA] support for 3D [1,1,X] import/output

* [GNA] add check for C == 1 and H == 1

* [GNA] fix input dimensions restrictions for 3d

3 years ago[IE][VPU] Proposal stage: fixed check for optional 2nd output (#2980)
Alexey Ershov [Thu, 5 Nov 2020 16:05:37 +0000 (19:05 +0300)]
[IE][VPU] Proposal stage: fixed check for optional 2nd output (#2980)

Fixed check for omitted 2nd optional output

3 years agoFixed disable/enable logic in PassConfig (#2940)
Gleb Kazantaev [Thu, 5 Nov 2020 14:34:32 +0000 (17:34 +0300)]
Fixed disable/enable logic in PassConfig (#2940)

* Fixed disable/enable logic in PassConfig

* Removed set_pass_config method for Manager; added comments

3 years agosetupvars.bat: Fixed names for major and minor python version vars (#2986)
Artyom Anokhov [Thu, 5 Nov 2020 11:48:09 +0000 (14:48 +0300)]
setupvars.bat: Fixed names for major and minor python version vars (#2986)

3 years ago[install_prerequisites.sh] remove libgfortran5, add framework to venv folder name...
Alexey Suhov [Thu, 5 Nov 2020 11:19:47 +0000 (14:19 +0300)]
[install_prerequisites.sh] remove libgfortran5, add framework to venv folder name (#2969)

3 years agoFix ngraph core opset createinsensitive (#2961)
Emmanuel Attia [Thu, 5 Nov 2020 10:42:09 +0000 (11:42 +0100)]
Fix ngraph core opset createinsensitive (#2961)

* Fix: ITT_INCLUDE_DIR was not correctly detected

* Map find result not compared with the right iterator.

* Force recheck

3 years ago[IE][VPU] Fix NMS DTS (#2880)
Andrew Bakalin [Thu, 5 Nov 2020 10:33:16 +0000 (13:33 +0300)]
[IE][VPU] Fix NMS DTS (#2880)

Add a new constructor to fix absent NMS-5 inputs that will be introduced after #2450 will be merged.

3 years agoUpdate ONNX models preprocessing script (#2936)
Rafal Blaczkowski [Thu, 5 Nov 2020 09:40:48 +0000 (10:40 +0100)]
Update ONNX models preprocessing script (#2936)

* Update onnx zoo models

* Update xfail list

* Update xfails

3 years agoMoved find_package(TBB) inside set_ie_threading_interface_for (#2966)
Ilya Lavrenov [Thu, 5 Nov 2020 08:32:46 +0000 (11:32 +0300)]
Moved find_package(TBB) inside set_ie_threading_interface_for (#2966)

3 years agoEnable LTO only for Release (#2968)
Ilya Lavrenov [Thu, 5 Nov 2020 08:31:32 +0000 (11:31 +0300)]
Enable LTO only for Release (#2968)

3 years agoMulti process build for MSVC (#2974)
Ilya Lavrenov [Thu, 5 Nov 2020 08:31:02 +0000 (11:31 +0300)]
Multi process build for MSVC (#2974)

3 years agoUpdate xfails (#2979)
Rafal Blaczkowski [Thu, 5 Nov 2020 07:12:51 +0000 (08:12 +0100)]
Update xfails (#2979)

3 years agoRemove obsoleted v0::Sum op (#2938)
Mateusz Tabaka [Thu, 5 Nov 2020 06:13:12 +0000 (07:13 +0100)]
Remove obsoleted v0::Sum op (#2938)

3 years agoRemove obsoleted v0::Slice and v0::Split ops (#2908)
Mateusz Tabaka [Thu, 5 Nov 2020 06:11:45 +0000 (07:11 +0100)]
Remove obsoleted v0::Slice and v0::Split ops (#2908)

* Remove obsoleted v0::Slice op

* Remove deprecated v0::Split op

* Fix build_graph tests

3 years agoRemove obsoleted v0::QuantizedConvolution op (#2958)
Mateusz Tabaka [Thu, 5 Nov 2020 06:10:51 +0000 (07:10 +0100)]
Remove obsoleted v0::QuantizedConvolution op (#2958)

3 years agoFix testcase CheckExecGraphInfoSerialization (#2973)
Jozef Daniecki [Thu, 5 Nov 2020 04:16:24 +0000 (05:16 +0100)]
Fix testcase CheckExecGraphInfoSerialization (#2973)

* Generate unique output file names in CheckExecGraphInfoSerialization testcase.

When multiple instances of this test were executed in parallel the same
file was accessed by multiple threads which was the cause of flakiness.

* Enable ExecGraphTests.CheckExecGraphInfoSerialization on GPU

3 years agoUpdate Custom_Layers_Guide.md (#2964)
Zoe Cayetano [Thu, 5 Nov 2020 04:01:21 +0000 (20:01 -0800)]
Update Custom_Layers_Guide.md (#2964)

Update misspelling on ONNX

3 years agoFix serialization of execution graph. (#2901)
Jozef Daniecki [Thu, 5 Nov 2020 03:52:08 +0000 (04:52 +0100)]
Fix serialization of execution graph. (#2901)

* Fix serialization of execution graph.

* Add exec graph comparison.

* Align execution graph serialization to old aproach.

* Revise error massages.

* Fixed execution graph comparison.

Now only attribute names are compared since values can differ on
different devices.

* Readability refactoring.

* Refactoring regarding readability.

3 years ago[IE CLDNN] Added inference precision into execution graph (#2913)
Vladimir Paramuzov [Wed, 4 Nov 2020 21:13:49 +0000 (00:13 +0300)]
[IE CLDNN] Added inference precision into execution graph (#2913)

3 years ago[CPU] CTCLoss sporadic issue. (#2934)
Nikolay Shchegolev [Wed, 4 Nov 2020 18:16:53 +0000 (21:16 +0300)]
[CPU] CTCLoss sporadic issue. (#2934)

* [CPU] CTCLoss sporadic issue.

* [Tests] Enable CTCLoss tests.

3 years ago[IE CLDNN] Add asymmetric quantization support to fsv16 imad general convolution...
Jedrzej Hajduczenia [Wed, 4 Nov 2020 14:31:40 +0000 (15:31 +0100)]
[IE CLDNN] Add asymmetric quantization support to fsv16 imad general convolution kernel (#2778)

3 years ago[IE CLDNN] Convolutition optimizations for small channels count (#2849)
Sergey Shlyapnikov [Wed, 4 Nov 2020 11:41:12 +0000 (14:41 +0300)]
[IE CLDNN] Convolutition optimizations for small channels count (#2849)

3 years ago[ONNX][PyTests] Fix result datatype missmatch (#2596)
Tomasz Socha [Wed, 4 Nov 2020 11:19:40 +0000 (12:19 +0100)]
[ONNX][PyTests] Fix result datatype missmatch (#2596)

3 years agoExpand ONNX functions to sub-graphs before import (#2733)
Michał Karzyński [Wed, 4 Nov 2020 09:48:34 +0000 (10:48 +0100)]
Expand ONNX functions to sub-graphs before import (#2733)

Co-authored-by: Bartosz Sledz <bartosz.sledz@intel.com>
3 years agoAzure CL: Improve yaml (#2951)
Alexander Zhogov [Tue, 3 Nov 2020 18:20:53 +0000 (21:20 +0300)]
Azure CL: Improve yaml (#2951)

* Azure CL: Improve yaml

* Fix installing requirements

* Add run onnxImporterUnitTests

3 years ago[IE] Add batched blob support (#2203)
Rafik Saliev [Tue, 3 Nov 2020 18:19:26 +0000 (19:19 +0100)]
[IE] Add batched blob support (#2203)

* [IE] Add batched blob support

New `class BatchedBlob : public CompoundBlob` defined to allow to pass multiple blobs as 1 InferRequest input.

Motivation: There is the special user case when a number of plain images (e.g. `NV12Blob`) should be passed as one input for network which batch size > 1.

`class CompoundBlob` is not applicable for such cases due to:
1. `NV12Blob` is `CompoundBlob` which prevents to combine multiple NV12 images to a CompoundBlob
2. The default behavior in most of plugins - do not accept generic CompoundBlob as `SetBlob()` argument

Adding `SetBlob(name, vector<Blob::Ptr>...)` to `class IInferRequest`, `class InferRequest`, `class IInferRequestInternal`, ...  - is not effective solution due to limited and specific use cases for `batched inputs`.

+ Apply rule-of-zero to CompoundBlob and inherited classes.

* Add "BATCHED_BLOB" optimization capability metric

* Add BatchedBlob usage to hello_nv12_input_classification

* Apply offline code review outcome:

1. Revert CompoundBlob public .ctors signatures
2. Remove 'workaround' .ctor for `BatchedBlob`
3. Revert tensor descriptors of `I420Blob` `NV12Blob` back to the 'fake' value.

* Code review fix

* Add functional tests for CPU, GPU, MULTI, HETERO

* update doc comment

* Apply code review change requests.

3 years agoUpdated requirements dependency (#2950)
Evgeny Lazarev [Tue, 3 Nov 2020 15:26:11 +0000 (18:26 +0300)]
Updated requirements dependency (#2950)

3 years agoSigmoid revise (#2942)
Piotr Szmelczynski [Tue, 3 Nov 2020 15:16:16 +0000 (16:16 +0100)]
Sigmoid revise (#2942)

* remove sigmoid_backprop

* Update Sigmoid spec

* Update Sigmoid spec

3 years ago[IE TESTS] Fix reg exp for issue 42029 (#2957)
Irina Efode [Tue, 3 Nov 2020 15:12:24 +0000 (18:12 +0300)]
[IE TESTS] Fix reg exp for issue 42029 (#2957)