* Revise reference implementation for ReduceMax operation
* Refactor backend unit tests
* Move tests with zero dims to op_eval
* Remove test with double elem type
* Fix code style
* Added minor changes
* Replace CoordinateTransform for CoordinateTransformBasic
* Added constant expression to set keep_dims as false
* Add const qualifier to local variables
* Use host tensor to retrieve and normalize axes
* StridedSlice spec refactored against explicit type indication.
* Add name to data input.
* Add new examples.
* Changed T to 'any supported type'.
* Remove mention about 'generalized python indexing' from short
description.
* refactor part of the docs file to use \dots
* refector docs
* add function enclosure for docs
* split function enclosurs across lines
* add latex operations to spec
* fix style
* fix missing index
* remove link to tensorflow operation
* Remove commas from formula.
Co-authored-by: jdanieck <jozef.daniecki@intel.com>
* Remove CoordinateTransform call to index function to calculate tensor element indexes
* Allow negative axis values in axes host tensor
* Added constant expression to set keep_dims as false
* Use rank from host tensor to normalize axes
* Address minor comments
* Add const qualifier to local variables
* Add deprecated macro for arm plugin dependent function signatures
* Remove duplicate helper functions
* Revise reference implementation for ReduceL1 operation
* Revise reference implementation for ReduceL2 operation
* Move op_eval tests to backend unit tests
* Added minor changes
* Replace CoordinateTransform for CoordinateTransformBasic
* Added constant expression to set keep_dims as false
* Add const qualifier to local variables
* Use rank from host tensor to normalize axes
* Create output shape for Convoution Backprop SLTs.
* Add output_padding attribute to SLT scope.
* Introduce SLT for Serializaton.
* Introduce new test layer class ConvolutionBackpropLayerTest which contains output_padding attribute and output_shape input. Old one is deprecated, but cannot be removed due to kmb plugin dependency.
* Add ConvolutionBackpropDataLayerTest into TEST_P.
* ConvolutionBackpropDataLayerTest left as legacy class used by kmb_plugin.
* Remove redundant variables.
* Switch to new API for gpu SLTs.
* Remove legacy API.
* Introduce legacy API to match dependency for KMB and ARM plugins.
* Create test cases for output_padding attribute.
* Fixing smoke_Deconv tests.
* remove FusedOp from selu operator
* add type_prop tests for selu op
* add vistors test for selu op
* update CMakeLists and fix wrong path to visitors test for selu
* fix style
* add backend tests for selu op
* refactor validate_and_infer_types function for selu op
* refactor type_prop tests for selu to catch particular errors
Co-authored-by: jdanieck <jozef.daniecki@intel.com>
* Revise reduction operations base class
* Fix visitor tests for logical reduction ops
* Remove node validation checks that break backward compatibility with tests
* Check for integer element type of axes input
* Check for unique elements provided in axes input
* Set mvn fusion test to have axes input with element type integral for ReduceMean nodes
* [LPT] Revise reduction operations base classes support in LPT
* Add node validation checks for integer element type of axes input in reduction operations
* Add end of line to files
* Implement output shape inference of reduction ops as member function
* Add ReduceBase class to have a common method to infer output shape
Co-authored-by: Edward Shogulin <edward.shogulin@intel.com>
* Revise reference implementation for ReduceSum operation
* Refactor backend unit tests
* Move tests with zero dims and dynamic shapes to op_eval
* Removed tests for f64 precision not supported
* Add support for precisions bf16, i8 and u8
* Clean up interpreter manifest unit tests
* Create ticket for result mismatch and accuracy failing tests
* Disable back interpreter tests failing for ubuntu18 i386 for large tensors
* Revert "Add support for precisions bf16, i8 and u8"
This reverts commit 8f1bba0dff.
* Add to manifest tests with disable types of reference implementation
* Address review comments
* Use constexpr variable to set keep dims as false in reduction output shape
* Replace CoordinateTransform with CoordinateTransformBasic
* Fix typo in variable name
* Remove keep_dims parameter from sum function call
* [GNA] Convert Matmul with batch size > 8 to pointwise convolution.
Support Eltwise split to more than 2 parts.
Fake Quantize support fixes.
* Put convolution split into a separate transformation
* Add separate transformations for cases with bias and fake quantize
* Rollback restriction for diagonal layer reshape
* Add transformation to normalize negative indices in Gather
* Update transformation name and add constant indices input to the pattern
* Apply comments
* Fix copyright year
* Add more tests
* Update type logic
* Add test with different types and update old ones
* Update pattern and logic to check only axis dimension is static, add appropriate tests
* Fix cdestyle
* Add axis normalization
* Fix wrong value in gather input
* Add related tests
* Add axis_constant to check
* Remove ngraph_check
