Skip to content

Fp8 gemm tweak for 405B Decoding #3104

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 1 commit into from

Conversation

zjing14
Copy link
Contributor

@zjing14 zjing14 commented Sep 10, 2024

Summary:
Improve fp8 gemm for memory-bound cases from 405B decoding

  • improved 405B decoding GEMMs:
    • [1, 13312, 16384] from 74us to 71us
    • [1, 16384, 6656] from 34us to 31us
  • Adjust tuning parameters that improves memory load performance
  • Adjust pipeline for overlapping scaling and gemm

Differential Revision: D62363038

Summary:
Improve fp8 gemm for memory-bound cases from 405B decoding
- improved 405B decoding GEMMs:
  - [1, 13312, 16384] from 74us to 71us
  - [1, 16384, 6656] from 34us to 31us
- Adjust tuning parameters that improves memory load performance
- Adjust pipeline for overlapping scaling and gemm

Differential Revision: D62363038
Copy link

netlify bot commented Sep 10, 2024

Deploy Preview for pytorch-fbgemm-docs ready!

Name Link
🔨 Latest commit ce0e85a
🔍 Latest deploy log https://app.netlify.com/sites/pytorch-fbgemm-docs/deploys/66dfa74b2f3c800008424bf4
😎 Deploy Preview https://deploy-preview-3104--pytorch-fbgemm-docs.netlify.app
📱 Preview on mobile
Toggle QR Code...

QR Code

Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify site configuration.

@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D62363038

@facebook-github-bot
Copy link
Contributor

This pull request has been merged in 04d3c58.

q10 pushed a commit to q10/FBGEMM that referenced this pull request Apr 10, 2025
Summary:
Pull Request resolved: facebookresearch/FBGEMM#192

X-link: pytorch#3104

Improve fp8 gemm for memory-bound cases from 405B decoding
- improved 405B decoding GEMMs:
  - [1, 13312, 16384] from 74us to 71us
  - [1, 16384, 6656] from 34us to 31us
- Adjust tuning parameters that improves memory load performance
- Adjust pipeline for overlapping scaling and gemm

Reviewed By: jianyuh, jwfromm

Differential Revision: D62363038

fbshipit-source-id: 4beea7eb8c66605539f9dd675a61ba3fe8870a3f
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants