* Moved stress tests to OV API 2.0
* Fix for using ouput index instead of get_index
* Removed ov::runtime namespace in stress tests
* Updated stress tests according to latest changes in OV 2.0
* Fix memleaks tests
* Updated run_memcheck.py to process gtest_filter
* Updated fillTensors, added InferAPI1 and InferAPI2 classes
* Updated test_inference_with_streams
* Updated isImage, comments
* Updated fillTensors to fill image_info inputs with positive pseudo-random numbers
* Removed redundant variable in fillTensors
* Update AUTO OV 2.0 c++ configuration API.
Signed-off-by: ywang2 <yang4.wang@intel.com>
* Support the OV 2.0 key to set model priority and add the test case to verify the prioirty map logic within AUTO plugin.
Signed-off-by: ywang2 <yang4.wang@intel.com>
* Replace the old model priority key and add the corresponding test case.
Signed-off-by: ywang2 <yang4.wang@intel.com>
* Fix LSTMSequence/GPUSequence validation behavior consistent with RNNSequence
Fixed issue with no exception if num_directions=2, but 'm_direction' is not set to BIDIRECTIONAL. Previously there was no error with this (and luckily it failed later in some CPU transformations during compile_network)
Corrected several tests which use copy-pasted num_directions=2 without m_direction set
Also for dynamic 'num_directions' - output shape still has 1 or 2 directions, because m_direction is known. Tests for GRU/LSTM are updated for this
Also several tests worked incorrectly for LSTMv0 - expectation was specific error to be thrown, but no expection was also allowed
* Fixed clang-format
* Store the expected output data in the TestCase class
* Skip the failing ONNX If tests
* Disable failing ONNX Softmax tests
* Disable the remaining failures
* ROI tensor support for Template plugin + tests for Template and CPU plugins
GPU doesn'tsupport ROI tensors, so tests were not added for GPU
* Added asserts for unsupported mixed axis order (like 0,3,1,2), and unsupported types like int4/int2 for ROI tensors
* Avoid duplicated outputs with the same name
* Revert onnx graph changes
* Allow output duplicates in ie_plugin_internal check
* Add test with onnx model
* Check get_tensor_ptr instead of any_name
* More outputs test
* Refactor to use std::transform
* test manifest update
* Remove redundant header
* INTERPRETER segfaults fix for duplicated output names
* Simplify duplication assert
* Update test names
* Test update
Currently, calling QueryNetwork from Myriad plugin with dynamic network could result in exception, this PR should fix this by removing nodes that could cause it from consideration.
Co-authored-by: Polina <polina.brzezinskaya@intel.com>
Support matmuls with two non-const inputs.
Detect concat inputs to matmul as changing batch size and
handle appropriately.
Enable tests in GNA_SW_EXACT mode for convolution stride > kernel size.
* Remove fp16 of Convert layer test from skip_tests.config.cpp as it works now
* update repo
* initial code commit
* add runtime reference
* apply ov::Model
* initial lstmcell-1 definition
* initial change
* apply Peepholes
* apply input_forget option
* apply initial test case of lstmsequence-1
* fix clang-format error
* fix clang-format error 2
* add lstms_sequence test cases by runtime reference and onnx test cases
* fix clang-format error
* fix clang-format error
* fix onnx test failure of LSTM IE_CPU
* fix clang-format issue
* fix clang-format issue 2
* add type_prop and visitor api test of lstm_sequence_v1
* fix clang-format error
* replace input/refOut data to hard coded and remove unnecessary enum definition
* update namespace of Tensor()
* remove supported test cases in disabling list
This PR fixes error
inference-engine/src/vpu/graph_transformer/src/stages/interpolate.cpp:65 Current Interpolate does not support resize by channels that started to appear after sample refactoring.
Issues: -75837
* [GNA] Fix large eltwise split in case split dimension is less than alignment
* Update src/plugins/intel_gna/optimizer/gna_pass_manager.cpp
Co-authored-by: Krzysztof Bruniecki <krzysztof.bruniecki@intel.com>
Co-authored-by: Krzysztof Bruniecki <krzysztof.bruniecki@intel.com>
* Further fixes of plugins.xml generation
1) Unregistration is done by name (e.g. CPU), not by file name (ov_cpu_plugin)
2) Unregistered line is searched by name="MULTI" instead of just 'MULTI' to not conflict with MULTI_WORK_MODE_AS_AUTO entry
3) Removed list of all possible plugins from ov_runtime as logic shall not rely on this (not possible to add 3rd party plugins)
* Revert ov_runtime - some CI jobs require plugins.xml even though plugins are not built
Registration - if some entry already exists in XML - don't copy it.
E.g.
- Registration of 'TEMPLATE' is performed
- Registration loops through existing plugins.xml
- If name="TEMPLATE" is found - don't take it to newContent
- If name like "myCustomPlugin" is found - take it
- As result - "myCustomPlugin" will exist after update, but old "TEMPLATE" will be removed
* Add missing change
* Detect casting node inside preprocessing block and leave it
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix unit-test for ObjectDetectionAPI
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Implement interpolate-1 and interpolate-4 in template plugin test
* Fix clang error
* Fix clang error
* Fix clang error
* Fix linux build
* Fix cpplint error
* Fix clang error
* Fix linux build
* Change way to pass attribute into struct for linux build
* Correct supported type
* Fix clang error
* Add visitor api test for interpolate-1 and interpoate-4
* Fix typo and cpplint
* Update copyright
* Avoid using a class that will be deprecated
* Rollback to CoordinateTransform
* Remove interpolate.in.cpp
Without fix, build of 'cpuUnitTests' will fail as 'inference_engine_s' will contain:
IE_STATIC_DEPENDENT_FILES = file_utils.cpp
$<TARGET_OBJECTS:${TARGET_NAME}_obj> - containing 'unity_cxx' which includes 'file_utils.cpp'
This causes multiple definition error of all methods inside file_utils.cpp
- Added registration (and unregistration) of ov_auto_batch_plugin. Otherwise 'BATCH' plugin will always produce new XML line without removing old one
- Added unregistration of legacy plugin names (<= 2021.4 release). Otherwise old lines like "libHeteroPlugin.so" will not be removed from plugins.xml file
* grouped convolution resolver fix
* add reverse_infer for Transpose, Elementwise, Identity and some fixes to successfully OOB convert MXNet models
* add reverse_infer for BatchNorms, dynamic groups fix, multi_box_prior.py fix
* add reverse infer for trivial cases
* input rank setting for image detection ops
* reverse_infer for Gather, AttributedGather add todo comment for Parameter
* reverse infer for LSTM, swapaxis; fixes and unit-tests for Elementwise, Gather, Transpose; also generalized Parameter reverse_infer; fixed DetectionOutput and partial reverse_infer
* clarified comment for LSTM reverse_infer
* detalized Elementwise reverse_infer, added several more unit-tests, Split reverse_infer fix, IteratorGetNext_ext.py fix
* generalized group convolution resolver
* specified in_ports explicitly
* clarified Split, Concat, Parameter, other minor typos
* relaxed compatibility check for elementwise; simplified scalar extracting for convolution.py and gather.py, fixed replacer order for ConvolutionNormalizer.py
* added routine set_input_shapes for image detection ops
* fixed typos to pass locally E2E, slightly changed LSTM reverse_infer
* VariadicSplit reverse_infer bugfix
* IteratorGetNextCut.py added check if port is connected
* explicitly specified in_port, added asserts, improved Split reverse infer
* sorted imports
* reverse_infer for ONNX and MXNet LSTM, GRU, RNN operations
* replaced Error raising with red warning
* added missing return for defined input shapes, fixed getting num groups for tf conv
* added mandatory attributes; fixed failing unit-tests, fixed typo with scalar group in ConvolutionNormalizer
* removed priorbox reverse_infer; Parameter reverse_infer corrected to use the only one existing out_port
* fixed clarify_partial_shape_test
* Preserve outputs order TF.
* Preserve of input/output indices for MxNet.
* Small fix.
* Added check.
* Small fix.
* Corrected Keras model importing.
* Fixed Keras model loading.
* Small correction.
* Corrected model loading.
* Small fix.
* Comment corrected.
* Removed unnecessary import.
* add performance hint to time infer
* add reshape methods for time infer tool
* add func declarations to common_utils.h
* add check if node is static (reshape pipeline)
* changes in desc of reshape funcs
* fix logic operator
* add reshape utils.h | move parsing logic in main func
* add tests_shared_lib to CMakeLists.txt
* change fill blobs with setStaticShapesBlobs
* change delimiter
* Preserving of input/output indices for ONNX.
* Fixed checks.
* Fixed for case of multiple outputs of node before Result.
* Added test for multiple tensor names before Result.
* Multiple tensor names before Result fix.
* Added order alignment for user input/output.
* Extended for case of input names in Parameter tensor list.
* Fixed unit tests.
* Corrected help.
* Small correction.
* Code refactoring.
* Temporarily reverted refactor.
* Fixed wrong changes.
* Fixed wrong changes.
* Returned reverted refactoring.
* Removed inputs_list from serializing.
* [Python API] Remove offline transformations from old python api
* [Python API] Move wheel folder to the python dir
* codestyle files
* [Python API] Move old python bindings
* move api
* one more codestyle
* fix install stage
* remove unexisting file from gitignore
* flake8 style
* try to fix jenkins
* undo comments
* return variable
* remove cpack
* fix after merge
* try to fix version
* fix test
* remove debug info
* version of python
* try to fix pyversion
* update tests
Co-authored-by: Alexander Zhogov <alexander.zhogov@intel.com>
* Loop/If/TensorIterator - fix dynamic input cases
Reference evaluate for body uses Model::evaluate instead of custom evaluation
Loop/TensorIterator additional fix - set result shape according to body execution result
Only op_eval test verifies issues, template tests were added just in case (these passed even without fix)
* Fix clang-format
* rename ti.cpp
* move gpu functional tests for range to correct location
* disable range layer test on gpu
it needs dynamic shapes support
Co-authored-by: Serhii Pavlovskyi <spavlovskyi@lohika.com>
* Squash commit: implement Conversion extensions
* Refactor PaddlePaddle FrontEnd
* Codestyle
* FrontEnd,InputModel,Place base classes -> abstract, renamed model file
* Fix unit tests
* fix unit tests
* ngraph:: to ov::
* Rename frontends dir to frontend
* fix merge conflicts
* Fix ConversionExtension
* get rid of NamedInputs/Outputs in TF FE
* Rename paddlepaddle to paddle; pdpd to paddle
* add missing file
* codestyle
* Remove local change
* paddlepaddle -> paddle for azure configs and .md files
* fix package name, fix config files
* Fix win build
* Revert Broadcast/AutoBroadcast changes
* codestyle
* fix FronEnd class
* fix ngraph_cpp_api.config
* fix incorrect merge, codestyle
* fix conversion extension
* conversion extension
* codestyle
* merge master
* fix build
* refactoring; revert broadcast/autobroadcast changes
* codestyle
* fix MacOS config
* resolve merge conflicts
* refactor includes
* register ConversionExtension in FrontEnds
* move get_op_type to base NodeContex class
* set op_translator map in ctor of Paddle FE; fix unit tests
* update unit tests; codestyle
* codestyle
* preliminary version of conversion extension in pybind
* conversion extension
* get_attribute_as_any method for NodeContext
* move get_attribute methods to NodeContext base class, rename get_ng_input to get_input
* add missed file
* Implement ov::Any getter in ONNX NodeContext
* fix py bindings
* Add update ConversionExtension unit tests, add SO unit tests, fix TF FE
* fix segfault on destructor
* fix NodeContext interface, fix unit tests
* set different names for ConversionExtensions in unit tests
* fix PaddleFuzzy tests
* fix Paddle Fuzzy tests
* revert changes in generate_slice.py
* fix codestyle
* fix pybindings
* revert local changes in generate_slice.py
* delete duplicate exceptions.hpp
* Refactoring: fix names according to convention
* pybinding for NodeContext, FrontEnd, ConversionExtension; fix unit tests; implement new unit tests
* Refactoring
* fix the case when a new converter rewrites existed one; delete unnecessary NodeContext from pybindings; use CreatorFunctons from the base class in ConversionExtension; update unit tests
* Revert local change
* PythonAPI: fix get_attribute method; fix get_input method; implement support of dtype and default attributes
* Fix py unit tests: add support for vector<ov::element::Type> as attribute
* resolve review comments
* fix unit tests
* move extension_holder to openvino/frontend/extension folder
* fix build on mac os
* temporary disable cast from vector<bool> to investigate issue on mac os
* Resolve review comments
* Resolve review comments
* Use dev API for .so extension
* Link frontends to pyopenvino as separate targets
* Temporary enable tf fe installation
* ignore PEP8 E402 for init files, set correct directory for py modules
* revert local changes
* Fix deadlock in pybind GIL; fix Win build; fix PEP8
* fix PEP8
* Add a return type annotation
* fix builds; fix ON/OFF switcher for ENABLE_OV_xxx_FRONTEND cmake options
* Fix the issue with ifdefs on WinOS; fix the issue related to pybindings and static c++ object
* fix python unit tests
* fix static build on WinOS
* Retrigger CI builds
* Fix static build on Windows
* fix static build on Windows again
* Retrigger CI
* delete unused includes; add a comment about issue on MacOS
* fix missprint
* resolve review comments
* fix missprint
* resolve review remarks
* Resolve review comments
* win win wheels build
* resolve review comments
* Python bindings - test for telemetry extension
This also ensures that actual 'Telemetry' object containing callbacks is still alive even there is no explicit Python objects holding it
* Fix pylint
* fix clang-format
* Remove fp16 of Convert layer test from skip_tests.config.cpp as it works now
* update repo
* add op reference test of region_tolo
* add type_prop test and remove backend test of region_yolo
* apply type conversion for loading file test and add bf16 test case in skip_test_config
* change location of compile definition under target and use path_join from file_util
* add dependency of test_model_zoo
* apply ov::Model
* remove unnecessary
* changed compile definition of TEST_FILES
* skip test cases of external test file
* remove test cases of importing data file
* Use std::string for static map instead of py::str
Probable reason is that this static map is destroyed after 'pybind' module is destoryed itself, thus py::str can't be cleaned up properly
* Added test via 'subprocess' execution of separate file
* if only one Device, not select
Signed-off-by: Hu, Yuan2 <yuan2.hu@intel.com>
* modify test case to match logic
Signed-off-by: Hu, Yuan2 <yuan2.hu@intel.com>
* [PYTHON] Add python APIs for loadNetwork and compile_networt without device name
CVS: https://jira.devtools.intel.com/browse/CVS-75249
Change-Id: Ia28e35f4ee66fc8fc5997b5bafe1b159670f9a21
Signed-off-by: River,Li <river.li@intel.com>
* Fix clang issue
Change-Id: I9988b16863af0e3883e99369f124cd05761d3210
* Fixed positional arguments issue
Change-Id: I6c3aa98bb693a619fa54fd6e96cf5eb89cdb9369
* Fixed 2 blank lines issue
Change-Id: I7f2afd7ebb80867a69d0c3ac9a6d4a38d95edb12
* Set AUTO as default device if no device name is set
Change-Id: Ic8646b12af0a2ab2fec6a07f5a12d460dcf781d7
* Resolve comments from code reviewer
Change-Id: Ia47faeb48937096e41e22ac59fbd88ec82cc6733
* moving the HETERO logic to the Auto-Batch (WIP), reverting to the ALLOW_AUTO_BATCHING and using that in the GPU remote tests
* shortned the vars names in the ie_core and prevented recursive auto-batching calls by checking for exclusive requests and disabling further auto-batching in the plugin, when HETERO is involved
* checking for the batch-dim presence (this is still WA until the https://github.com/openvinotoolkit/openvino/pull/9559 is merged) - pls see CVS-75317
+clang for the ie_core.cpp
* moving the HETERO logic back to the ie_core.cpp, storing the _so internally for no-batch code-path
* [IE SAMPLES] activated NCC tool for c++ samples
* exclude ov_ncc_naming_style for tests
* fixed NCC hit
* Added support for source files in samples
* changed style of methods for benchmark
* changed style for speech sample
* changed code style
* changed code style for shared_tensor
* benchmark changes
* changed remote_tensors_filling
* fixed notes
* rebase of branch
Useful in cases when pad_value is a subgraph marked as decompression
which cannot be constant folded, but still, Pad can be eliminated
if pad_value evaluates to zero.
* fix: set shape in fbc if undefined dimension in model
* Update tools/pot/openvino/tools/pot/algorithms/quantization/fast_bias_correction/algorithm.py
Co-authored-by: Liubov Talamanova <piccione-mail@yandex.ru>
* Apply suggestions from code review
Co-authored-by: Liubov Talamanova <piccione-mail@yandex.ru>
* update shape in bc id dynamic shape
Co-authored-by: Liubov Talamanova <piccione-mail@yandex.ru>
* changed permutes
* fixed permutes
* fixed kernel
* fix transpose after convolution
* fix for convnet
* insert transposes for all convolutions and poolings
* refactor transformations;
added unit tests;
removed old transformations for addinf permutes/reshapes
* fixed constant types
* fixes after merge
* fixed bug for rm_cnn4a: added correct time_dim for the first convolution
* added fix for timeheightconvolution: in this case we have correct time set in convolution kernel already
* minor review fixes: renamed transformation and file
* rename in test
* rename in test
* sort imports + couplt changes in comments
* review fixes: refactoring
* replaced recursive implementation by nx.topological_sort;
fixed comments
* minor fixes: comment + preserving node names
* Graph comparator - take sinks into accounts
Previously graph has been traversed only from Results, so any differences in 'Sinks' were not detected
* Fix functional tests
* Update after internal discussion
* Fix low_latency_test (addition to low_latency_v2_test)
* Fix typo
* add limit format on snprintf
Signed-off-by: Hu, Yuan2 <yuan2.hu@intel.com>
* add limit on format
Signed-off-by: Hu, Yuan2 <yuan2.hu@intel.com>
* add test case
Signed-off-by: Hu, Yuan2 <yuan2.hu@intel.com>
* fix a bug for LOG_TRACE
Signed-off-by: Hu, Yuan2 <yuan2.hu@intel.com>
* remove debug info
* Change dynamic_to_static to dinamic_to_undefined function to use -1 values instead of upperbounds in serialized IRs
* Resolve conflict after src files transition
* Remove resolve_dynamic_shapes function
* Fix typo
* Fix codestyle
* add warning about order if both mean and scale set
* Update tools/mo/openvino/tools/mo/main.py
Co-authored-by: Anastasia Popova <anastasia.popova@intel.com>
* removed warning, added phrase in documentation
* fixed merge
* added phrase about order of ,mean and scale in MO help
* duplicate MO help phrase in doc
* Update docs/MO_DG/prepare_model/convert_model/Converting_Model.md
Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>
* Update docs/MO_DG/prepare_model/convert_model/Converting_Model.md
Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>
* Update docs/MO_DG/prepare_model/convert_model/Converting_Model.md
Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>
* Update tools/mo/openvino/tools/mo/utils/cli_parser.py
Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>
* Update tools/mo/openvino/tools/mo/utils/cli_parser.py
Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>
* remove tabs
* fix in order of reverse, mean, scale
Co-authored-by: Anastasia Popova <anastasia.popova@intel.com>
Co-authored-by: Anastasiya Ageeva <anastasiya.ageeva@intel.com>
* [Python API] Rename offline_transformations in the new api
* remove comments
* one more place to rename
* fix mo building
* fix after merge
* fix for pot import
* Add include guard to file_utils.cpp
* Rebase src
* Rename acc tests, fix rebase
* Revert debug changes
* Fix linter
* Move ac tests to new template
* Test updated
* Fix result operation are sharing output tensor with previous op
* Pruning test visualzation option
* Add ac support for all test cases with pruned kernels
* Remove redundant files
* Enable extended pruning logging by env variable
* Adjust pruning tests
* Remove pruning extended debug env var
* Enable init masks only in debug mode
* Set result mask to input tensor instead of output tensor by separate key
* Bug fix / Test coverage
* Fix comments
* passing by reference to fix performance inefficiencies.
Signed-off-by: Wang, Yang <yang4.wang@intel.com>
* Passing by const reference.
Signed-off-by: Wang, Yang <yang4.wang@intel.com>
* Q/DQ + mulichannel support
backup
fix interval
mfk_functiun.cpp
WIP moveDequantizationBefore
add moveDequantizationBefore function
add cpu and gpu tests
attribute cmp false
attribute cmp false
rm temp line
mkl-dnn update
concat with multichanels for mOve_fake_quantize_function, bad runtime info for q/dq
rm extra qualification
fix run time info for q/dq
add support of multichanel fakequantize, bad test for it
work tests for multi chanel FQ
rm workaround
cpplint fix
cpplint fix
don't worl Variadic split
ieFuncTest work
cpuFuncTest work
Fix benchmark_app build (#7577)
[GPU] Added onednn dependency. (#6564)
cpp lint
cpplint
fix get_shape
fix fq constants
cpp lint
some fix in mfk.cpp
resolve conversations, add spil_nodes function
add new tests for multi-chanels, rename NetworkHelper::split_consts_before_concat()
fix get fq constants
* add new multi-chanels test and use constant_fold to split constant
* remove extra spaces
fix namespase terminated
fix namespase terminated
* resolved minor comments
* added check for convert_q
* Eliminate no-op elementwise operations
This change adds EliminateEltwise pass to NopElimination.
EliminateEltwise removes:
- Subtract with zero
- Multiply with one
- Divide by one
* add Add support to EliminateEltwise
* fix unit test
* use get_all_data_elements_bitwise_identical instead of get_single_value
* fix are_all_data_elements_bitwise_identical for constant created from HostTensor
* fix lpt tests
* check for mul in is_dequantization_subgraph function
* optimize fetching constant value
* apply review comments
* [Common FrontEnd] Implement ChangePlaceholderType transformation with OldApiMap
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Add ChangePlaceholderType pass into frontend normalize
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Add a comment about the transformation
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix build issue
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Apply code review comments: remove ngraph namespace rountine, etc.
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Remove use of extra Node object in set_old_api_map call
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Create separate pass for legacy nGraph transformations for MO with new frontend
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix build for ieFuncTests
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix build issue with namespace conflicts
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Revert "Fix build issue with namespace conflicts"
This reverts commit 50a189f4e5.
* Move legacy transformations to ngraph namespace
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix build issue with unresolved MOCLegacyTransformation
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Adjust code to new OldApiMap API
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Resolve merge conflicts
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix build issue
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Do not set legacy type for parameter in case user defined type
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Removed old dynamism API from infer request
* Fixed segfault
* Fixed some tests
* Added method to set inputs, outputs
* Fix python tests
* Try to fix window build
* Calculate model layout based on 'tensor' layout and convert steps
Previously, 'model layout' is set to '...' by default,
thus no shape conversion happened when tensor layout is set to 'NHWC', then there was explicit convert_layout "NCHW"
Now "model layout" is calculated based on tensor layout and conversion steps:
Examples:
1) Tensor: NHWC, Convert: NCHW. Result: NCHW
2) Tensor: NHWC, Convert: 0312. Result: NCHW
* Initial move of tensor data calculation
* Moved 'impls' to new file
* Postprocessing + unit tests
* clang-format fix
* Added more details to preprocessing nodes
- Mean/Scale - will print mean/scale values
- Convert type - will print type
- Convert layout - will print destination layout
- Convert color - will print destination color
It is needed to troubleshoot the problems. If error occurs, message will not display last op's target shape/layout/type
* Add python bindings
* update tests
* Added memory type to dump if set
* Code style fix
* unity build fix
* Dump tensor if only memory type is set
* Added debug print
* Fix Param->Result case
Previously, layout was set by preprocessing set to old parameter as well
This is incorrect because in case of exception layout info will not be reverted
In this case old Result pointed to old Parameter and was able to preserve runtime info
After fixing of this, case Param->Result was broken if revalidation is not triggerred
Fix is to detect 'Result' as a consumer of some parameter and force revalidation in this case
* Revert occasionally committed line
* And one more line
* rename inference_engine to OpenVINO
* correct exception for batch
* check all inputs to find batch dimension before throwing exception
* correct warning about batch
* avoid set_shape in static case
* refactoring latency output
* message about benchmarking mode
* use new precision naming
* use pass manager instead offline_transformations
* Move 'NV12toRGB/BGR' reference evaluates to template plugin
CPU doesn't need this fallback, so implementation can be moved to reduce core binary size
* Moved evaluate_nv12 to 'runtime::reference'
* Fix arm build
* ShutdownProtobufLibrary when unload paddle frontend dynmaic library to fix probuf memory leak
* ShutdownProtobufLibrary if the frontend libraries use protobuf
* make shutdown_protobuf a library
* Set THROUGHPUT as the default configration for all the plugin and display the config of the plugin.
Signed-off-by: Wang, Yang <yang4.wang@intel.com>
* updated format.
Signed-off-by: Wang, Yang <yang4.wang@intel.com>
* Update benchmark python API.
Signed-off-by: Wang, Yang <yang4.wang@intel.com>
* Replace str 'THROUGHPUT' with CONFIG_VALUE(THROUGHPUT).
Signed-off-by: Wang, Yang <yang4.wang@intel.com>
* Using CONFIG_VALUE(THROUGHPUT) replace 'THROUGHPUT' string.
Signed-off-by: Wang, Yang <yang4.wang@intel.com>
* update code style.
Signed-off-by: Wang, Yang <yang4.wang@intel.com>
* Move the setting output code into the try block.
Signed-off-by: Wang, Yang <yang4.wang@intel.com>
* Calculate model layout based on 'tensor' layout and convert steps
Previously, 'model layout' is set to '...' by default,
thus no shape conversion happened when tensor layout is set to 'NHWC', then there was explicit convert_layout "NCHW"
Now "model layout" is calculated based on tensor layout and conversion steps:
Examples:
1) Tensor: NHWC, Convert: NCHW. Result: NCHW
2) Tensor: NHWC, Convert: 0312. Result: NCHW
* Fix for set_shape + resize case
* Implement the batch to space shape infer
* Implement the space_to_batch shape inference.
* Implement shape infer of space_to_depth and depth_to_space OPs
* Fix Azure building issue.
* Add namespace for the shape_infer function.
* Avoid using friend declaration for shape infer.
* update coding style issue
* Update based on review comments
* Apply review comments
* Add test cases.
* Update the shape infer flow.
* Fix the bug in the previous test case.
* Update coding style.
* Fix the code bug caused by the DepthToSpace check fix.
* update coding style.
* Implment the Dimension/StaticDimension division operator by a value
* Refine the the code.
* Fix the issue when T is implicitly construct StaticShape with PartialShape when comparing
* Update the CI issue.
* Move the shape_infer helper into src folder.
* Apply the review comments.
* Coding style fix.
* Remove the ngraph folder
* Applied review comments
* Fix CI windows building issue
* Move test into new folder.
* Not support divisor is negative.
* Apply review comments.
* Fix CI issues
* Apply review comments.
* Update
Co-authored-by: Evgenya Stepyreva <evgenya.stepyreva@intel.com>
* Q/DQ + mulichannel support
backup
fix interval
mfk_functiun.cpp
WIP moveDequantizationBefore
add moveDequantizationBefore function
add cpu and gpu tests
attribute cmp false
attribute cmp false
rm temp line
mkl-dnn update
concat with multichanels for mOve_fake_quantize_function, bad runtime info for q/dq
rm extra qualification
fix run time info for q/dq
add support of multichanel fakequantize, bad test for it
work tests for multi chanel FQ
rm workaround
cpplint fix
cpplint fix
don't worl Variadic split
ieFuncTest work
cpuFuncTest work
Fix benchmark_app build (#7577)
[GPU] Added onednn dependency. (#6564)
cpp lint
cpplint
fix get_shape
fix fq constants
cpp lint
some fix in mfk.cpp
resolve conversations, add spil_nodes function
add new tests for multi-chanels, rename NetworkHelper::split_consts_before_concat()
fix get fq constants
* add new multi-chanels test and use constant_fold to split constant
* remove extra spaces
fix namespase terminated
fix namespase terminated
* Updated requirements for MO and POT with telemetry.
* Added mock telemetry common class for unit tests.
* Used mock telemetry in preprocessing unit tests.
* Small correction.
* Fix in the transformation PreserveRuntimeInfo: now Transpose is inserted before input port 0 of Result only, not after data node of layer before Result layer.
* Deleted commented code.
* Added more tests for the MO transformation PreserveRuntimeInfo.
* Use fp16-int8 mixed precision, instead of fp32-int8 mixed precision for onednn
* Allow quantization fusion into bsv32_fsv16 conv
* For conv, do not select bsv16_fsv16. Select bsv32_fsv16 for mixed-layout
* depthwise conv is supported even though it is not fp16
* Allow resample kernel to work as cross-layout
* test case for cross-layout of resample_opt kernel
* Select onednn-friendly format from cldnn conv
* Optimization for fp16 mixed precision
* Choose mixed layout in case of mixed precision from reorder_inputs
* Support for mixed precision from depth_to_space
* Do not convert first conv format
* Use onednn for FC output of fp16
* Choose bsv8_fsv4 from quantization even when conv kernel size is not 7
* Select cldnn for first conv when input feature depth is 1
* For first conv, use onednn only when kernel size is 7x7
* Use short variable name and added is_i8_u8 helper function
Co-authored-by: Kim,SungEun <sungeun.kim@intel.com>
* [LPT] Documentation
* 1) ToC was removed 2) SVG => PNG temporary conversion
* [LPT] Refactoring + developer guide
* [LPT] attribute doxygen documentation was added
* [LPT] Developer Guide to Reference API links were added
* [LPT] comments fixes
* [LPT] Reference API to Developer Guide links were added
* [LPT] titles were changed
* [LPT] comments fixes#2
* [LPT] root document was moved to Plugin DG
* [LPT] Documentation: image link quick fix
* [LPT] Docummentation: PrecisionsAttribute description quick fix
* fix comments from Karol
* fixes
* movement
* directive was added
* movement #2
* LPT reference in Executable Network rollback
* snippets were updated ini accordance with new API
* Handle names collisions for old IR with new API
* Fixed load model
* Try to fix tests
* Try to fix tests
* Try to fix build
* Try to fix tests
* Fixed tests
* Revert "Fixed tests"
This reverts commit 35da307210.
* Refactoring
* Fixed functional test
* Try to fix CPU tests
Co-authored-by: Ilya Lavrenov <ilya.lavrenov@intel.com>
* [Python API] Remove offline transformations from old python api
* try to fix import error
* try to fix pylint
* try to fix pylint2
* Use new api in graph_utils
* Fix pylint
* Try to fix pylint
* Use serialize from pass manager
* try to skip tests
* try to use new ir
Co-authored-by: AlexeyLebedev1 <alexey.lebedev@intel.com>
* [GPU] Enable unet2d enable on DG2
Add to support is_os_yx_isa2_osa8_isv8_osv2 format, which is used in
weight reorder.
Signed-off-by: hyunback <hyunback.kim@intel.com>
* [GPU] Enable implicit concat batch1 in oneDNN.
* Use gpu_usm memory offset, enable implicit concat batch1 in oneDNN.
And optimized_out node doesn't always have to be mutable input,
so add to check whether mutable input is existed in optimized node.
* Update to check use_usm condition in implicit concat.
* Add the condition for implicit concat.
* implicit concat's dependency should not be fused_op with eltwise.
* Buffer reuse is required for onednn sum post operation, output padding
did the buffer reuse failure.
Signed-off-by: hyunback <hyunback.kim@intel.com>
+ cldnn supports hard sigmoid activation function but onednn doesn't.
+ split it into eltwise linear and eltwise clip in
add_onednn_optimization_attributes pass.
* Add MatMulMultiplyFusion
MatMulMultiplyFusion replaces following subgraph:
MatMul->Multiply (with const)
to following:
Multiply->MatMul
where Multiply is applied to MatMul's second input.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.