* Fix errors in VariadicSplit layer restored from serialized IR
* Update VariadicSplit specification and error message to allow 1D tensors on 1st input
* Update spec
* Resolve comments
* Apply comments, add unit tests
* Update unit tests
Original source code repo: https://github.com/llohse/libnpy
SHA of original commit: d3fd88697889cefb466b647d3034c1d7b7f615ff
In OpenVINO repo there are some modifications, thus Intel's copyrights are kept as well
* Swap inputs pass
* [GNA] Handle case Gemm layer
* [GNA] Convert Matmul to FC
* VS tests
* Move to common optimization
* Gemm execution such as FC
* Test calc scale_factor
* Some changes
* Working version
* [GNA] Insert transpose between convolution/pooling and reshape.
Insert copy layers after concat inputs with multiple connections to the concat.
Accept networks with input connected to layers with different orientations if one of input dimensions is 1.
Fix scale factor calculation for Eltwise layer.
Fixes for Gemm quantization.
* Insert transpose after Reshape and before Matmul
* Fix concat input alignment when it's the network input
* Comments applying
Co-authored-by: Andrey Dmitriev <andrey.dmitriev@intel.com>
* Added LoadNetwork(filename) to AUTO
* Added more files
* So pointer can be used without loading
* Changed InferencePlugin, ICore to return internal interfaces
* Added SoPointers for InferRequest, ExecutableNetwork
* Fixed Windows
* Fixed KMB
* Fixes for KMB
* Removed dereference operator
* Play with include files
* Fixed compilation with older compilers
* Fixed comments
* Fixed win build
* Try to fix Windows
* Try to fix Windows 2
* Fixed windows
* Fixed windows
* Removed SOPointer as a base class
* Reverted back SOPointer split
* Code review
Co-authored-by: apankratovantonp <anton.pankratov@intel.com>
* [IE TESTS] Fix comparation in LayerTestUtils
* Fixes
* Small fix
* Int4 fixes
* remove extra
* Fix NMS
* Some fixes for tests
* Add small fix
* [IE TESTS] Remove const folding as a result engine
* Remove extra
* Revert remove constant folding (DSR test) & fix some cases for cpu
* Fix GNA
* add conversion of padded to valid convolution without other parameters change
* [GNA] Fix graph loop when multiple connections exist from single layer to concat
* [GNA] Add 1d and 2d conv test cases
Add models covering all transform scenarios.
Add test cases covering 1d and 2d convolutions.
Update transform with the newest code.
Add minor fixes in transform and elsewhere.
* [GNA] Remove debug code
* [GNA] Fixes after review
* [GNA] Fix failing tests
Co-authored-by: prozen <piotr.rozen@intel.com>
* Implement nGraph transformation to decompose Einsum-7 operation
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Use MatMul instead of Eltwise-multiplication and ReduceSum
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Add description for new methods
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix code style
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix code style #2
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Remove unused variables.py
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Apply feedback after review: fix comments, new_register_node use
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Add Reshape if needed and apply code-review feedback
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Fix code-style
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Remove unused variable
Signed-off-by: Roman Kazantsev <roman.kazantsev@intel.com>
* Remove extern template from headers for RTTI classes
* Moove instantiation out of the namespace
* Use __ANDROID__ conditional compilation for TBlob
* One more attempt
* Reference implementation for memory
* memory reference implementation tests, fixes
* new class Variable context
* fix ngraph code style
* add new evaluate method to ngraph::function
* unordered_set instead of set in find_variables method
* added a new Memory base class; automatiс memory allocation for Variable context in Assign ops; refactoring
* ngraph codestyle
* ngraph codestyle
* temporary disable the check of variables in ngraph::function
* fix for evaluate hides overloaded virtual function warning
* ngraph codestyle
* uncomment a check in validate_and_infer method
* Removing a check (not backward compatible); adding docs
* Auto detect Parameters/Variables, new constructors in ngraph::function
* use zero initial values in ReadValue v6 evaluate
* fix unit tests
* fix codestyle
* fix build (werror)
* ngraph codestyle
* update unit tests, refactoring
* ngraph codestyle
* refactoring, docs, new unit tests
* Resolve review remarks
* rt_attributes likeapproach in EvaluationContext, codestyle
* fix build and unit tests
* resolve review comments
* resolve review comments
* codestyle
* export public API
* Add axis support
* Update dequant extractor
* Update qdeq ops
* Refactoring quantize
* Update dequantize resolver
* Update dequantize op
* Refactoring dequantize
* Some fixes for quantize and dequantize
* Update unit tests
* Reafctoring quantize/dequantize axis support
* Move quantize/dequantize resolvers to middle
* hot fix
* Fix unit tests
* Fix unit tests
* Update quintize resolver comment
* Refactoring code according to code review
* Fix according to review
* Change order for transforms quantize pipline
* Python API for LoadNetwork by model file name
* BenchmarkApp: Add caching and LoadNetworkFromFile support
2 new options are introduced
- cache_dir <dir> - enables models caching
- load_from_file - use new perform "LoadNetwork" by model file name
Using both parameters will achieve maximum performance of read/load network on startup
Tests:
1) Run "benchmark_app -h". Help will display 2 new options. After available devices there will be list of devices with cache support
2) ./benchmark_app -d CPU -i <model.xml> -load_from_file
Verify that some test steps are skipped (related to ReadNetwork, re-shaping etc)
3) Pre-requisite: support of caching shall be enabled for Template plugin
./benchmark_app -d TEMPLATE -i <model.onnx> -load_from_file -cache_dir someDir
Verify that "someDir" is created and generated blob is available
Run again, verify that loading works as well (should be faster as it will not load onnx model)
4) Run same test as (3), but without -load_from_file option. Verify that cache is properly created
For some devices loadNetwork time shall be improved when cache is available
* Removed additional timing prints
* Correction from old code
* Revert "Removed additional timing prints"
Additional change - when .blob is chosen instead of .xml, it takes priority over caching flags
* Removed new time printings
As discussed, these time measurements like 'total first inference time' will be available in 'timeTests' scripts
* Fix clang-format issues
* Store weights range in meta info instead of cloning whole constant
* Add command line option for constants size serialization threshold
* Update IR Runner to handle OP meta information
* Fix interpolate type_prop tests
* Skip new failures in SLT
* Fix models count
* Add dynamism elimination option.
* TopK shape propagation changed
* Fix type_prop tests for TopK
* Update specification for ConvolutionBackpropData.
* Add backticks to attribute types, changed layout description for input, filter and output.
* Correct xml example.
* Add new examples.
* Add link with convolution backprop description.
* Repleace additional link with argxiv website.
* Insert enumeration for examples.
* Fix example with output_shape input.
* add upgrading gather1 -> gather7 transformation
* added G1->G7 to common_optimizations list but by default in turned off
* fixed a couple of typos in comments
* corrected validation error messages for GatherBase
* removed redundant comments
* clang format fix
* coNverts v1::Gather into v7::Gather
* added explicit batch_dims = 0, corrected axis bound check for dynamic data_rank
* Add visitor test for convolution_backprop.
* Add test to CMakeLists, corrected input shapes.
* Add checking attributes count.
* Extend test for all autoPad types.
* Support old TBBs
* Don't reset environment
* Removed useless NO_CMAKE_FIND_ROOT_PATH
* Fixed build with old TBB for Windows
* Fixed ngraph code style
