Skip to content

Enable TP for Mamba-based models #14811

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.

Sign up for GitHub

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

Merged
Fridge003 merged 2 commits into from
Dec 11, 2025
+0 −1
Merged

Enable TP for Mamba-based models #14811

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
040c9f4
Enable TP for Mamba-based models
roikoren755 Dec 10, 2025
9d1c3bd
Merge branch 'main' into feat/nemotronh-tp-fix
Fridge003 Dec 10, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 0 additions & 1 deletion python/sglang/srt/layers/attention/mamba/mamba.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -339,7 +339,6 @@ def __init__(
input_is_parallel=True,
quant_config=quant_config,
prefix=f"{prefix}.out_proj",
reduce_results=False,
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Dec 10, 2025

Choose a reason for hiding this comment

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

critical

By removing reduce_results=False, the RowParallelLinear layer will use its default behavior of reduce_results=True. This is the correct fix, as it enables the all-reduce operation needed to combine the partial outputs from each tensor parallel rank. Without this, the output would be incomplete and incorrect when TP > 1, which explains the garbage output mentioned in the pull request description.

)

self.norm = Mixer2RMSNormGated(
Expand Down
Loading