Open
Conversation
…ht-tying / grad-zeroing / etc
…n a mask, a hook and a custom autograd function
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adding 3 different weight freezing approaches in an example script under:
notebooks/embedding_wrapper.py.The first wrapper uses a mask and ended up being equivalent to the code that Chris already had set up.
The second wrapper uses a hook. That's only recommended for debugging, so ultimately not useful for us. It helped me spot an error in the other code though, so I think the pattern is useful and I'm keeping it in for reference.
The third wrapper uses a custom autograd function. That may be interesting, since it's more lowlevel than the masking code.
Finally, this code contains a test setup which doesn't pull in hf / transformers/ etc. That's convenient for quick debugging.