Enable flag to not pass PAD tokens in ffwd #775
Merged
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
bcui19
changed the title
vchiley
reviewed
Dec 4, 2023
vchiley
reviewed
Dec 4, 2023
Can you run main vs your branch with use_pad_tok_in_ffwd flag vs your branch without use_pad_tok_in_ffwd flag? |
dakinggg
reviewed
Dec 4, 2023
dakinggg
reviewed
Dec 5, 2023
vchiley
reviewed
Dec 5, 2023
Co-authored-by: Mihir Patel <mihir.v.patel7@gmail.com>
Co-authored-by: Mihir Patel <mihir.v.patel7@gmail.com>
dakinggg
reviewed
Dec 8, 2023
vchiley
reviewed
Dec 11, 2023
vchiley
reviewed
Dec 11, 2023
mvpatel2000
approved these changes
Dec 11, 2023
vchiley
approved these changes
Dec 11, 2023
Co-authored-by: Daniel King <43149077+dakinggg@users.noreply.github.com>
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Outdated suggestions cannot be applied.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.
This PR does two things:
attn_biasfunction to always return the attention_mask..forwardon the ffwd network then re-add in the pad tokens.Loss curves on a fully randomly initialized network:
We also get slightly higher throughput from this when there are PAD tokens in our dataset (and no degradation when compared to
mainwithattn_impl: triton:wandb: https://wandb.ai/mosaic-ml/padding_check?workspace=user-bcui