fix: prefer inplace softmax to avoid copy #2661
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
With chunking in place, we don't want to increase max-batch-prefill-tokens anymore, but instead choose an optimal value :). That being said, nice find, probably a lot of opportunities for such updates (let's focus on the large tensors though) |
Narsil
previously approved these changes
Oct 17, 2024
Narsil
reviewed
Oct 17, 2024
Co-authored-by: Nicolas Patry <patry.nicolas@protonmail.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Outdated suggestions cannot be applied.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 modifies
log_softmaxto operate in place, eliminating the need to copy large tensors. This optimization reduces memory consumption during warmup.For instance, when using
meta-llama/Meta-Llama-3.1-8B-Instructon an L4, this change allows running the model with--max-batch-prefill-tokensincreased from 7192 to 9874 without exceeding memory limits