fix the gradient backward issue when joint training with s3prl frontend #5159
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
simpleoier
force-pushed
the
joint_train_fix
branch
from
5c62225
to
33aa097
Compare
simpleoier
force-pushed
the
joint_train_fix
branch
from
3434492
to
d3dbd76
Compare
Codecov Report
@@ Coverage Diff @@
## master #5159 +/- ##
==========================================
- Coverage 74.99% 74.99% -0.01%
==========================================
Files 618 618
Lines 55588 55589 +1
==========================================
Hits 41689 41689
- Misses 13899 13900 +1
Flags with carried forward coverage won't be shown. Click here to find out more.
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
|
@Emrys365, can you review this PR? |
Emrys365
approved these changes
May 20, 2023
Comment on lines
-54
to
-60
| if getattr( | ||
| upstream.upstream, "model", None | ||
| ) is not None and upstream.upstream.model.__class__.__name__ in [ | ||
| "Wav2Vec2Model", | ||
| "HubertModel", | ||
| ]: | ||
| upstream.upstream.model.encoder.layerdrop = 0.0 |
There was a problem hiding this comment.
Why are these lines removed?
There was a problem hiding this comment.
I think this is because S3PRL already sets encoder_layerdrop=0 when initializing an upstream, so we no longer need to do this in ESPnet. Am I right?
There was a problem hiding this comment.
Yes, it is correct.
Emrys365
approved these changes
May 20, 2023
|
Thanks a lot for fixing it! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In s3prl,
feature_grad_multwas set to 0. Thus the forward is in the context oftorch.no_grad()here. It will stop the gradient back-propagation in joint training. To fix it, just set it to be 1 manually.Also in this PR, the encoder layerdrop part is removed, because it is set to 0 from s3prl for all upstreams, e.g. WavLM.