Skip to content

[Sequence parallel][train] Support sequence parallelism without sample packing#480

Merged
SumanthRH merged 5 commits intoNovaSky-AI:mainfrom
SumanthRH:seq-parallel-no-pack
Oct 15, 2025
Merged

[Sequence parallel][train] Support sequence parallelism without sample packing#480
SumanthRH merged 5 commits intoNovaSky-AI:mainfrom
SumanthRH:seq-parallel-no-pack

Conversation

@SumanthRH
Copy link
Member

No description provided.

Signed-off-by: SumanthRH <sumanthrh99@gmail.com>
Signed-off-by: SumanthRH <sumanthrh99@gmail.com>
x
Signed-off-by: SumanthRH <sumanthrh99@gmail.com>
Signed-off-by: SumanthRH <sumanthrh99@gmail.com>
@SumanthRH SumanthRH marked this pull request as ready for review October 15, 2025 00:41
@SumanthRH
Copy link
Member Author

/gemini review

Copy link
Contributor

@gemini-code-assist gemini-code-assist bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request successfully extends sequence parallelism to support operation without sample packing. The changes are logical and well-implemented across the affected files. Key modifications include generalizing padding and slicing utilities to handle batch sizes greater than one, updating the model wrapper to correctly manage attention masks and gather outputs in both packed and non-packed scenarios, and adding a necessary re-computation of query_length for flash attention. The accompanying test enhancements are thorough, covering both with and without sample packing cases. I have one minor suggestion to update a docstring for consistency.

x
Signed-off-by: SumanthRH <sumanthrh99@gmail.com>
Copy link
Collaborator

@erictang000 erictang000 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm

@SumanthRH SumanthRH merged commit 9f1dac4 into NovaSky-AI:main Oct 15, 2025
3 checks passed
li-boxuan pushed a commit to li-boxuan/SkyRL that referenced this pull request Nov 23, 2025
dzorlu pushed a commit to fleet-ai/SkyRL that referenced this pull request Feb 4, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants