Revise LRN reference implementation (#2672)
authorMateusz Tabaka <mateusz.tabaka@intel.com>
Mon, 19 Oct 2020 05:40:04 +0000 (07:40 +0200)
committerGitHub <noreply@github.com>
Mon, 19 Oct 2020 05:40:04 +0000 (08:40 +0300)
commit5965010bec08da471abe1da7495e87a6d7b3f6da
tree82346f24708639bd5878e3c99201a352b39d8121
parent84b5fc51dcb63f3f83e8ed8954d32bc979da4f71
Revise LRN reference implementation (#2672)

* fix typo in LRN docs

* fix link to reference in LRN doc

* LRN, LRN_IE types alignment with spec

* align LRN ref implementation to plugins behavior

* update LRN docs

* Improve LRN reference implementation performance

* restore LRN constructor with no axes in the input

* apply code format

* revert double->float size_t->int change

* small fix to example in doc

* revert double->float size_t->int in onnx_importer and backend tests

* Changes to docs after review
docs/ops/normalization/LRN_1.md
inference-engine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/lrn.cpp
inference-engine/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/lrn.cpp
ngraph/core/reference/include/ngraph/runtime/reference/lrn.hpp
ngraph/test/backend/lrn.in.cpp
ngraph/test/ref_generators/generate_lrn_across_axes.py
ngraph/test/runtime/ie/unit_test.manifest