Add initial ring flash attention support #1266
Draft
Conversation
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
Contributor
dmahan93 commented Sep 4, 2024
dmahan93 commented - Only tested with BF16 Zero2 (on a separate repo, so beware copy/paste errors)
- Made an assumption that SP should have preference over TP to be on the same node in the topology, but happy to revisit that
- Only supports the zigzag convention of ring attention
- The attention mask stuff is v important, it lets you do 128k context on 1 node even without ring attention
Contributor Author
| @Quentin-Anthony probably merge conflicts and other stuff, it's v untested, but it sounded like it would be better to have something than wait around so it's a bit rushed so forgive any annoying issues 😅 |
This was referenced Sep 6, 2024
Closed
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 Outdated suggestions cannot be applied.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.