* fake quantize single layer test for GNA plugin
* implemented fakequantize for fp32 case as an activation function
* added proper seed randomisation within single test run
* [GNA] [FAKEQUANTIZE] fixed ref-fp32 implementation on GNA to use nearbyint instead of roundf
* [GNA] [FAKEQUANTIZE] restored random seed
* [GNA][FAKEQUANTIZE] disabled 4d and integer tests for FakeQuantize
* [GNA][FAKEQUANTIZE]updated ngraph FakeQuantize builder to accept seed
* [GNA][FAKEQUANTIZE]aligned FP calculations order on GNA with reference ngraph - this however gives more error
* [CPU]build of FakeQuantise tests restored
* [TESTS][FAKEQUANTIZE] ignore extra inferRequests for disabled tests
* [GNA] Fixed legacy unit test failuers appeared due to extra check for possible segfault in import frames
* [GNA] adopted fuse multiple identities for FakeQunatize layer
* [GNA]fp32 runtime code review
* Specify in and out precisions separately, add layouts for convolution
* Align convolution layer tests instantiations with updated definition
* Align convolution layer tests instantiations with updated definition for template plugin
* net, in, out prcs
Co-authored-by: Mikhail Treskin <mikhail.treskin@intel.com>
some plugins
- added shared parameterized tests
- instantiated for template plugin
- instantiated for cpu plugin
- fixed CPU plugin to properly handle U16 input
- fixed CPU reverse_sequence primitive to alolw input/oputput tensors to
be in FP32 only
- updated ngraph test_simple_computation_on_ndarrays to not expect
failure on U16 input
* [GNA] fix scale factor calculation for unfused bias after fc
* change check
* add test
* apply requested changes
* cpplint fix
* apply test changes
* modify model for test to match ::op::
* Fix for concat layer with more than 2 inputs
Signed-off-by: Bartosz Sochacki <bartosz.sochacki@intel.com>
* Fixed check if affine is used for crop layer
Signed-off-by: Bartosz Sochacki <bartosz.sochacki@intel.com>
* code cleanup for fix affine layer check
Signed-off-by: Bartosz Sochacki <bartosz.sochacki@intel.com>
* added test for concat layer with multiple inputs
* simplified test to use less number of layers
* fixed code style
* fixed coding style
* addressed review comments and one more issue that appeared during testing
* fixed code style errors
* scale factor propagation for concat layer with multiple inputs
* fix for a case when all inputs to concat are activation layers
* fix for linux compilation - C++14 is not enabled and fails on lambda with auto parameters
* corrected current year in headers in concat multi input tests
* fixes for code review issues raised by Denis Orlov
* enabled integer mode computation in GNA concat multi input test
* removed 1 space per review comment
* a fix to fail when not all scale factors are equal
* added GNA_DEVICE_MODE config to concat multi input test
* corrected searching for a next input to concat layer
* changed selection of 2nd candidate for source quant value
* code style fix - else and brackets should be in the same line
* small code improvement
* fix for mixing line endings
* addressed with endless requantization loop and fixed failing tests
Special test case with input values which cannot be correctly processed via
decomposition with int AVG pool layer.
Signed-off-by: Alexander Peskov <alexander.peskov@intel.com>
* [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
[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
* 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.
* Removed reference implementations from public API
* Remove coordinate_transform from public API
* Introduced static library with reference implementations
* Added new predicates for smart pattern matching
* Added ConvMul and GroupConvMul fusion passes based on opset4; Added CPU functional tests for comparing fusion accuracy
* Improved ConvMultiply fusion to support scalars; Added positive and negative tests
* Added ConvolutionBackprop/GrouConvolutionBackprop Multiply fusion; Added functional tests
* [CPU] Add support 4th and 5th input DetectionOutput
* fix any comments
* move reference to ngraph
* some changes for mx nms
* change namespace for ref impl
* Implement unicode conversion using Windows native functions
* NOCPPLINT
* Fixed deprecated c++ api usage in tests
* Moved impl to cpp
* Moved Unicode utils to Plugin API
* Added missed include for Windows
* Fixes for unit tests; CentOS fixes
* Fixed Windows compilation
* Fixed unit tests on Unix
* Fixed unix 2
* Network serializer for v7 is removed
* Fixed compilation
* Fixed Windows build
* WA for GPU
* Create function 2 times
* Fixed compilation
* Added return
In some networks, mvTensor would request a large CMX-DMA transfer (<512K). That starves DMA for other timing critical tasks such as SIPP. Limit CMX-DMA request size as an option in myriad_compile:
* Add compile option TILING_CMX_LIMIT_KB
* Declare compile option TILING_CMX_LIMIT_KB in IE tools (compile_tool and vpu_compile)
* Add tests for compile option TILING_CMX_LIMIT_KB. Small fix for naming behavior tests.
* [LPT] FuseFakeQuantizeAndScaleShift transformation for last layer fix
* [LPT] refactoring
* [LPT] FuseFakeQuantizeAndScaleShift test: last layer name validation was added
* Remove replacement of StridedSlice with other stages and execute it on device as one kernel.
* Refactor strided slice tests to be able to parametrize it by precision.
* Update firmware.
* added support for power layer with non-1 exponents to GNA plugin
* reverted a change caused by merge issue
* fixes for review comments (typo fix - lrelu instead of leru, unnamed structure instead of of named one in union with arguments of activation function, name fix - input instead of inputs),
scale-shift implementation based on affine layer instead of PWL,
* fixed code style
* fixes for coding style in scale_factor_calc.hpp
* added domain for power function
* fixed review comment - power function specific methods
* added check if dynamic casting was successful
* removed I16 as it is not supported by ngraph
* fixed initialization per review comment
[GNA] Added fix multiple output with one go to memory and test
[GNA] Added fix multiple output with one go to memory and test
[GNA] Added fix multiple output with one go to memory and test
Added multi output
Update gna_pass_manager.cpp
test
[GNA] Added fix multiple output with one go to memory and test
[GNA] Added fix multiple output with one go to memory and test
[GNA] Added fix multiple output with one go to memory and test
Added multi output
Update gna_pass_manager.cpp
test
tests
[GNA] Added fix multiple output with one go to memory and test
[GNA] Added fix multiple output with one go to memory and test
Added multi output
Update gna_pass_manager.cpp
test
tests
Added pass
Test
test
tests_2
return old
* Execution graph via ngraph for CPU plugin
* Fixes
* Migrated to VariantImpl instead of Parameter
* Reverted to dedicated ExecutionNode once again
* Re-use new execution graph in tests
* Fixed one more tests to use execution graph via ngraph::Function
Anton Pankratv