Adding a fix for Cross Entropy Loss for long sequence lengths. #795
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
Pulling the latest commits from main fork
Pulling from the main repo
Pulling from mosaicml/llm-foundry main
Merging from mosaic main
Pulling from mosaic main
Pulling from mosaic main.
ShashankMosaicML
requested review from
sashaDoubov,
vchiley,
mvpatel2000,
dakinggg and
nik-mosaic
vchiley
reviewed
Dec 9, 2023
vchiley
approved these changes
Dec 10, 2023
dakinggg
approved these changes
Dec 12, 2023
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Outdated suggestions cannot be applied.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.
For longer sequence lengths (>16k), we see that the current cross entropy loss implementation throws an illegal memory access error. This PR fixes this by reverting to an older version of cross entropy loss that works for longer sequence lengths as well.
WandB link for the experiments: https://wandb.ai/mosaic-ml/longcont-CE-debug
We see the same convergence, mfu, and memory usage before and after the fix
