-
Notifications
You must be signed in to change notification settings - Fork 2.4k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
LSTM/GRU/RNN Sequences : support for seq_lengths input #2788
LSTM/GRU/RNN Sequences : support for seq_lengths input #2788
Conversation
itikhono
commented
Oct 23, 2020
•
edited
Loading
edited
- Sequences to TI transformations
- Adjust reference implementations of GRU/LSTM/RNN Seqs
- fix ngraph python API
- IR Reader unit tests
- enable ONNX importer unit tests
- fix TensorIterator ref impl and tests
…e reference implemetations, add new tests
…s_to_tensor_iterator
…s_to_tensor_iterator
…s_to_tensor_iterator
…s_to_tensor_iterator
...ngine/tests/functional/plugin/cpu/shared_tests_instances/single_layer_tests/gru_sequence.cpp
Show resolved
Hide resolved
...e/tests/functional/plugin/gpu/shared_tests_instances/single_layer_tests/reverse_sequence.cpp
Outdated
Show resolved
Hide resolved
.../transformations/src/transformations/op_conversions/convert_sequences_to_tensor_iterator.cpp
Outdated
Show resolved
Hide resolved
.../transformations/src/transformations/op_conversions/convert_sequences_to_tensor_iterator.cpp
Show resolved
Hide resolved
.../transformations/src/transformations/op_conversions/convert_sequences_to_tensor_iterator.cpp
Outdated
Show resolved
Hide resolved
body_params.push_back(aggregated_Y_h_body_param); | ||
|
||
// Create mask node deciding whether or not to mask batch data. | ||
ngraph::Output<ngraph::Node> batch_seq_length = ngraph::builder::opset1::legacy_broadcast_for_binary_operation( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If you create opset5 operations during transformation then you can't create opset1.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@GlebKazantaev opset5 ops are aliases to opset1, for all ops in this legacy_broadcast_for_binary_operation
function.
I don't want to duplicate the code of this function and don't want to upgrade the function to opset5 due to there are other places of use of this function.
This is a general problem in our code: we cannot create helper functions with a hard-coded version of the opset. It is a more complicated issue than using this function. I will open a question in our morning meeting.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
And we also allow MO to generate ops from different opsets
.../transformations/src/transformations/op_conversions/convert_sequences_to_tensor_iterator.cpp
Outdated
Show resolved
Hide resolved
.../transformations/src/transformations/op_conversions/convert_sequences_to_tensor_iterator.cpp
Show resolved
Hide resolved
.../transformations/src/transformations/op_conversions/convert_sequences_to_tensor_iterator.cpp
Outdated
Show resolved
Hide resolved
.../transformations/src/transformations/op_conversions/convert_sequences_to_tensor_iterator.cpp
Show resolved
Hide resolved
.../transformations/src/transformations/op_conversions/convert_sequences_to_tensor_iterator.cpp
Show resolved
Hide resolved
680e2cc
to
5014cf2
Compare
…s_to_tensor_iterator
@itikhono please resolve conflicts. |
…s_to_tensor_iterator
…s_to_tensor_iterator
…s_to_tensor_iterator
…it#2788) * sequences to ti transformations, support for seq_lengths input, update reference implemetations, add new tests * fix python api, update sequences to ti transformation * fix sequences to ti transformation * Update sequences to TI transformation: fix reverse sequence support * update single layer tests, fix TI reference impl, fix Sequences to TI transformations * ngraph code style * fix build * fix ngraph python api * resolver review comments, refactoring * Resolve review remarks * delete xfail
…it#2788) * sequences to ti transformations, support for seq_lengths input, update reference implemetations, add new tests * fix python api, update sequences to ti transformation * fix sequences to ti transformation * Update sequences to TI transformation: fix reverse sequence support * update single layer tests, fix TI reference impl, fix Sequences to TI transformations * ngraph code style * fix build * fix ngraph python api * resolver review comments, refactoring * Resolve review remarks * delete xfail