Skip to content

fix: fix fp8 _apply_state_dict_to_model#2233

Merged
terrykong merged 1 commit intomainfrom
yukih/fix-fp8-state-dict
Apr 8, 2026
Merged

fix: fix fp8 _apply_state_dict_to_model#2233
terrykong merged 1 commit intomainfrom
yukih/fix-fp8-state-dict

Conversation

@yuki-97
Copy link
Copy Markdown
Contributor

@yuki-97 yuki-97 commented Apr 8, 2026

  1. revert mistaken _apply_state_dict_to_model changes in feat: Add Eagle3 online speculative decoding support #2078.
  2. minor cleanup of get_pipeline_model_parallel_group and is_pipeline_last_stage.

fix the following tests:

  • grpo-llama3.1-8b-instruct-2n8g-megatron-fp8-e2e
  • grpo-moonlight-16ba3b-4n8g-megatron-fp8-e2e
  • grpo-deepseek-v3-64n8g-fp8-async-1off
  • grpo-llama3.1-8b-instruct-2n8g-fp8-async-1off

Signed-off-by: Yuki Huang <yukih@nvidia.com>
@yuki-97 yuki-97 requested review from a team as code owners April 8, 2026 10:41
@copy-pr-bot
Copy link
Copy Markdown

copy-pr-bot Bot commented Apr 8, 2026

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@yuki-97 yuki-97 added the CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version) label Apr 8, 2026
@yuki-97
Copy link
Copy Markdown
Contributor Author

yuki-97 commented Apr 8, 2026

/ok to test faf455a

@terrykong terrykong merged commit 12f3acc into main Apr 8, 2026
29 of 30 checks passed
@terrykong terrykong deleted the yukih/fix-fp8-state-dict branch April 8, 2026 22:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants