Skip to content
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.

Sign up for GitHub

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

Jump to bottom

[LPT][WIP] Support of INT8 inference of LSTM layer #10354

Closed
wants to merge 3 commits into from
Closed

[LPT][WIP] Support of INT8 inference of LSTM layer #10354

wants to merge 3 commits into from

Conversation

ndemashov
Copy link
Contributor

Details:

  • Support of INT8 inference of LSTM layer

Tickets:

  • 56479

@ndemashov ndemashov requested review from a team February 14, 2022 12:44
@openvino-pushbot openvino-pushbot added category: IE Tests OpenVINO Test: plugins and common category: CPU OpenVINO CPU plugin labels Feb 14, 2022
@dmitry-gorokhov dmitry-gorokhov added this to the 2022.2 milestone Feb 21, 2022
@ndemashov ndemashov force-pushed the nd/lpt/support_LSTM_INT8_inference branch 3 times, most recently from e0f2466 to 954485e Compare February 28, 2022 12:59
@ndemashov ndemashov force-pushed the nd/lpt/support_LSTM_INT8_inference branch 2 times, most recently from c5e572d to bfa8601 Compare March 15, 2022 12:19
@ndemashov ndemashov requested a review from a team as a code owner March 16, 2022 23:18
@ndemashov ndemashov requested review from a team and sbalandi and removed request for a team March 21, 2022 10:01
@ndemashov ndemashov force-pushed the nd/lpt/support_LSTM_INT8_inference branch 9 times, most recently from d8c12b8 to 822dd07 Compare March 28, 2022 22:55
@ndemashov ndemashov force-pushed the nd/lpt/support_LSTM_INT8_inference branch 3 times, most recently from db76daf to 4a9f4c7 Compare March 29, 2022 15:00
eshoguli
eshoguli reviewed Mar 29, 2022
View reviewed changes
Copy link
Contributor

@eshoguli eshoguli left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There are comments, tests review is still in progress.

src/common/low_precision_transformations/include/low_precision/lstm.hpp Outdated Show resolved Hide resolved
src/common/low_precision_transformations/include/low_precision/lstm.hpp Outdated Show resolved Hide resolved
...ommon/low_precision_transformations/include/low_precision/rt_info/skip_cleanup_attribute.hpp Show resolved Hide resolved
...ommon/low_precision_transformations/include/low_precision/rt_info/skip_cleanup_attribute.hpp Outdated Show resolved Hide resolved
...ommon/low_precision_transformations/include/low_precision/rt_info/skip_cleanup_attribute.hpp Outdated Show resolved Hide resolved
src/common/low_precision_transformations/src/lstm.cpp Outdated Show resolved Hide resolved
src/common/low_precision_transformations/src/lstm.cpp Outdated Show resolved Hide resolved
src/common/low_precision_transformations/src/lstm.cpp Outdated Show resolved Hide resolved
src/common/low_precision_transformations/src/lstm.cpp Outdated Show resolved Hide resolved
src/common/low_precision_transformations/src/lstm.cpp Outdated Show resolved Hide resolved
@ndemashov ndemashov closed this Mar 30, 2022
@ndemashov ndemashov deleted the nd/lpt/support_LSTM_INT8_inference branch March 30, 2022 18:44
@ndemashov ndemashov force-pushed the nd/lpt/support_LSTM_INT8_inference branch 3 times, most recently from e7e799b to c8ef697 Compare March 31, 2022 18:03
@ndemashov ndemashov requested review from a team March 31, 2022 21:10
@ndemashov ndemashov force-pushed the nd/lpt/support_LSTM_INT8_inference branch 5 times, most recently from a5222f1 to 6df5098 Compare April 3, 2022 09:25
eshoguli
eshoguli reviewed Apr 4, 2022
View reviewed changes
Copy link
Contributor

@eshoguli eshoguli left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

minor comments, please fix

...ommon/low_precision_transformations/include/low_precision/rt_info/skip_cleanup_attribute.hpp Show resolved Hide resolved
...ommon/low_precision_transformations/include/low_precision/rt_info/skip_cleanup_attribute.hpp Show resolved Hide resolved
...ommon/low_precision_transformations/include/low_precision/rt_info/skip_cleanup_attribute.hpp Outdated Show resolved Hide resolved
...ommon/low_precision_transformations/include/low_precision/rt_info/skip_cleanup_attribute.hpp Outdated Show resolved Hide resolved
src/common/low_precision_transformations/include/low_precision/network_helper.hpp Outdated Show resolved Hide resolved
src/common/low_precision_transformations/src/recurrent_cell.cpp Outdated Show resolved Hide resolved
src/common/low_precision_transformations/src/recurrent_cell.cpp Outdated Show resolved Hide resolved
@ndemashov ndemashov force-pushed the nd/lpt/support_LSTM_INT8_inference branch from 08ab512 to 90e33cf Compare April 4, 2022 10:33
eshoguli
eshoguli reviewed Apr 4, 2022
View reviewed changes
src/tests/functional/inference_engine/lp_transformations/recurrent_cell_transformation.cpp

const std::vector<std::vector<ngraph::Shape>> weights_shapes = {{{4, 1}, {4, 1}, {4}}};

const std::vector<RecurrentCellTransformationTestValues> testValues = {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Here and for plugin tests: add test case with zero-point's on activations, please.

src/common/low_precision_transformations/src/recurrent_cell.cpp Outdated Show resolved Hide resolved
@ndemashov ndemashov force-pushed the nd/lpt/support_LSTM_INT8_inference branch 4 times, most recently from 7634c44 to 8affc04 Compare April 5, 2022 08:46
ndemashov added 2 commits April 5, 2022 12:21
@ndemashov
LSTM transformation
5ef6df4 
added lstm_function and tests

fixed fake quantize on weights

added transformation

transformation for LSTMCell

added skip_cleanup_attribute

moved skip_cleanup_attribute to all dequantization operations & added GRUCell support

disabled constant folding on weights & tests refactoring

Only TensorIteratorCommon is supported by now

added lstm_sequence

removed extra template

fixed LSTMSequence matcher

added RNNCell node support

renamed transformation & refactoring

sk p_clean p_attribute transformation matchere  refactoring

added cpu plugin tests

added gpu plugin tests

added fakeQuantizeWraper

removed LSTMSequence

refactoring
@ndemashov
disabled plugins tests
e28d59a
@ndemashov ndemashov force-pushed the nd/lpt/support_LSTM_INT8_inference branch from 8affc04 to e28d59a Compare April 5, 2022 09:21
@ndemashov ndemashov marked this pull request as draft April 5, 2022 17:33
@dmitry-gorokhov dmitry-gorokhov modified the milestones: 2022.2, 2022.3 Aug 18, 2022
@dmitry-gorokhov
Copy link
Contributor

closed as outdated

@dmitry-gorokhov dmitry-gorokhov removed this from the 2022.3 milestone Nov 16, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@eshoguli eshoguli eshoguli left review comments

@sbalandi sbalandi Awaiting requested review from sbalandi

Assignees
No one assigned
Labels
category: CPU OpenVINO CPU plugin category: IE Tests OpenVINO Test: plugins and common
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@ndemashov @dmitry-gorokhov @eshoguli @openvino-pushbot