Skip to content

Add TE MXFP8 multi-stream implementation in TorchAO#3997

Draft
elfiegg wants to merge 1 commit intopytorch:mainfrom
elfiegg:te
Draft

Add TE MXFP8 multi-stream implementation in TorchAO#3997
elfiegg wants to merge 1 commit intopytorch:mainfrom
elfiegg:te

Conversation

@elfiegg
Copy link

@elfiegg elfiegg commented Mar 5, 2026

To-dos:

  1. Microbenchmark the kernel performance compared to TorchAO cutlass
  2. More holistic verification on workloads like DeepseekV3 and FLUX

Previously observed 1.24x speedup to e2e DeepseekV3 workload, see description: pytorch/torchtitan#2207

@pytorch-bot
Copy link

pytorch-bot bot commented Mar 5, 2026

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3997

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@meta-cla meta-cla bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Mar 5, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant