[GNA] Support in GNA plugin for power layer with non-1 exponents (#997)
authorBartosz Sochacki <bartosz.sochacki@intel.com>
Fri, 10 Jul 2020 10:39:29 +0000 (12:39 +0200)
committerGitHub <noreply@github.com>
Fri, 10 Jul 2020 10:39:29 +0000 (13:39 +0300)
commit8da662b2b80c340f2b08e2c32d6cb8a5f1b8f9ce
treef399015647a0726d4f2c359b45eced4f37a66681
parentd9706da8d0c205bf5d8c76a536cca10923326455
[GNA] Support in GNA plugin for power layer with non-1 exponents (#997)

* 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
inference-engine/src/gna_plugin/backend/am_intel_dnn.cpp
inference-engine/src/gna_plugin/backend/dnn_types.h
inference-engine/src/gna_plugin/backend/make_pwl.cpp
inference-engine/src/gna_plugin/frontend/scale_factor_calc.hpp
inference-engine/src/gna_plugin/gna_graph_compiler.cpp
inference-engine/src/gna_plugin/layers/gna_layer_info.hpp
inference-engine/src/gna_plugin/runtime/pwl.cpp
inference-engine/src/gna_plugin/runtime/pwl.h
inference-engine/tests/functional/plugin/gna/shared_tests_instances/single_layer_tests/power.cpp [new file with mode: 0644]
inference-engine/tests/functional/plugin/shared/include/single_layer_tests/power.hpp [new file with mode: 0644]
inference-engine/tests/functional/plugin/shared/src/single_layer_tests/power.cpp [new file with mode: 0644]