Change repeat to expand in GQA
#628
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
sashaDoubov
changed the title
expand to repeat in GQArepeat to expand in GQA
vchiley
approved these changes
Sep 26, 2023
|
Will this be backwards compatible with any MPT model using GQA, trained using the code from a commit before this? |
|
@madhavatreplit yes, this is backwards compatible, as the model weights don't change, this is simply an optimization for the attention function to avoid allocating new memory. |
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Outdated suggestions cannot be applied.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 reduces memory usage and improves MFU. Note that there may be some risk of NaNs with small d_head (this was previously seen for MQA). However, there is significant perf benefits for standard d_head.