Fine-tune FP8 BMM performance #3224
Closed
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
|
This pull request was exported from Phabricator. Differential Revision: D63882833 |
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
Summary: X-link: facebookresearch/FBGEMM#316 Marlin int4 weight-only with loopover for bmm performs great (**up to 7x faster** compared to bf16 bmm) when dim M is small to medium size (e.g., < 256) in decode; For larger dim M, we could leverage this bmm int4 rowwise weight-only kernel in prefill that is around **1.5x faster** than marlin int4 loopover and maintain the same accuracy More results can be found in this [data sheet](https://docs.google.com/spreadsheets/d/12JWt3SqX_1GSLKwjGyt0KQl9SMWDF0r0C63MMKsE9JM/edit?usp=sharing) Reviewed By: jianyuh Differential Revision: D63818529
Summary: X-link: facebookresearch/FBGEMM#322 Fine-tune FP8 BMM performance to get additional 20% performance gain. Differential Revision: D63882833
jiawenliu64
force-pushed
the
export-D63882833
branch
from
80e131d
to
e3458c2
Compare
|
This pull request was exported from Phabricator. Differential Revision: D63882833 |
|
This pull request has been merged in 7a4472a. |
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Outdated suggestions cannot be applied.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.
Summary: Fine-tune FP8 BMM performance to get additional 20% performance gain for given shapes
Differential Revision: D63882833