Skip to content

rocm: scaled_grouped_mm support gfx942 fp8 data type#3955

Merged
danielvegamyhre merged 1 commit into
pytorch:mainfrom
xiaobochen-amd:dev_fp8_gg_gfx942
Feb 27, 2026
Merged

rocm: scaled_grouped_mm support gfx942 fp8 data type#3955
danielvegamyhre merged 1 commit into
pytorch:mainfrom
xiaobochen-amd:dev_fp8_gg_gfx942

Conversation

@xiaobochen-amd

Copy link
Copy Markdown
Contributor

Old PR: #3540

  • Docker: rocm/primus:v26.1
  • Torch 2.10.0
  • ROCM 7.1
  • GPU: MI325X

Test Case:

  • [PASS] test/prototype/moe_training/test_training.py
  • [PASS] test/prototype/moe_training/test_scaled_grouped_mm.py

Benchmark:

  • python3 -m benchmarks.prototype.moe_training.benchmark_scaled_grouped_mm_dq --compile --profile
image

@pytorch-bot

pytorch-bot Bot commented Feb 26, 2026

Copy link
Copy Markdown

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3955

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

✅ You can merge normally! (1 Unrelated Failure)

As of commit 82d2499 with merge base cd062f2 (image):

BROKEN TRUNK - The following job failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

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 Feb 26, 2026
@danielvegamyhre danielvegamyhre self-requested a review February 26, 2026 18:32
@danielvegamyhre

Copy link
Copy Markdown
Contributor

test failure is unrelated, landing

@danielvegamyhre danielvegamyhre merged commit 4e18d87 into pytorch:main Feb 27, 2026
20 of 21 checks passed
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. topic: rocm

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants