Fix bidirectional mode in reference implementations of GRU/LSTM/RNN Sequences (#2264)
authorIvan Tikhonov <ivan.tikhonov@intel.com>
Fri, 18 Sep 2020 07:14:01 +0000 (10:14 +0300)
committerGitHub <noreply@github.com>
Fri, 18 Sep 2020 07:14:01 +0000 (10:14 +0300)
commit1b7dfc6e4c8efbac04a9d2dc57eca90641e387f1
treed499d29f01c07af5bb8fcc47127b035947da9d3a
parent93074590de9b58f380da6f6df02a856ac4fad99b
Fix bidirectional mode in reference implementations of GRU/LSTM/RNN Sequences (#2264)

* fix bidirectional case in references of sequences ops, enable decomposition of bidirectional cases in CommonOptimizations

* introduce new opset5, include GRU/RNN/LSTM Sequences to opset5

* Revert "introduce new opset5, include GRU/RNN/LSTM Sequences to opset5"

This reverts commit 73c22a11dbd724d2cfa9212ff211db74ef09cf2a.
inference-engine/src/transformations/src/transformations/bidirectional_sequences_decomposition.cpp
inference-engine/src/transformations/src/transformations/common_optimizations/common_optimizations.cpp
inference-engine/tests/functional/plugin/shared/src/single_layer_tests/gru_sequence.cpp
inference-engine/tests/functional/plugin/shared/src/single_layer_tests/lstm_sequence.cpp
inference-engine/tests/functional/plugin/shared/src/single_layer_tests/rnn_sequence.cpp
ngraph/core/reference/include/ngraph/runtime/reference/sequences.hpp