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.

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

[Onnx] Support Bidirectional RNNs #8337

Merged

Conversation

AndrewZhaoLuo
Copy link
Contributor

@AndrewZhaoLuo AndrewZhaoLuo commented Jun 25, 2021

LSTMs and GRUs now support the common bidirectional flag.

cc @mbrookhart

Copy link
Contributor

@mbrookhart mbrookhart left a comment

Choose a reason for hiding this comment

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

LGTM. Great work!

Copy link
Contributor

@jwfromm jwfromm left a comment

Choose a reason for hiding this comment

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

Very nice implementation, excellent PR @AndrewZhaoLuo!

@AndrewZhaoLuo AndrewZhaoLuo force-pushed the andrewluo-onnx-add-bidirectional-lstm branch from f1ce2bb to 804bc7b Compare June 28, 2021 06:31
@mbrookhart mbrookhart merged commit 2915349 into apache:main Jun 28, 2021
@mbrookhart
Copy link
Contributor

Thanks @AndrewZhaoLuo @jwfromm

ylc pushed a commit to ylc/tvm that referenced this pull request Sep 29, 2021
* modify lstm to be easily bidirectional

* make it obvious some matriciies are packed via prime notation

* fix var name

* more var names

* add op split

* keyword arg names

* missing implicit cls arg

* deal with extra dimensions

* last of the fixes

* refactor rnn tests to support directions

* bidirectional tests

* test forward results

* go backwards

* more fixes

* reverse tokens on reverse pass

* parameterized directions

* double up activations in bidirect

* slow attribute forgetting

* lstm interface is v. confus

* test forward complete

* add GRU outline

* revisiion2

* why was tehre a not

* gru tests

* missing bounds, copy pasta!

* add comment

* ensure all args fp
zxy844288792 pushed a commit to zxy844288792/tvm that referenced this pull request Mar 4, 2022
* modify lstm to be easily bidirectional

* make it obvious some matriciies are packed via prime notation

* fix var name

* more var names

* add op split

* keyword arg names

* missing implicit cls arg

* deal with extra dimensions

* last of the fixes

* refactor rnn tests to support directions

* bidirectional tests

* test forward results

* go backwards

* more fixes

* reverse tokens on reverse pass

* parameterized directions

* double up activations in bidirect

* slow attribute forgetting

* lstm interface is v. confus

* test forward complete

* add GRU outline

* revisiion2

* why was tehre a not

* gru tests

* missing bounds, copy pasta!

* add comment

* ensure all args fp
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants