Skip to content

Fix dtype error in CI #8197

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
ispobock merged 1 commit into from
Jul 20, 2025
Merged

Fix dtype error in CI #8197

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 1 addition & 1 deletion python/sglang/srt/layers/moe/topk.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -524,7 +524,7 @@ def biased_grouped_topk_gpu(
topk_weights = torch.empty((token, topk), dtype=torch.float32, device=device)
topk_ids = torch.empty((token, topk), dtype=torch.int32, device=device)
aiter_biased_grouped_topk(
gating_output,
gating_output.to(dtype=torch.float32),
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist bot Jul 20, 2025

Choose a reason for hiding this comment

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

medium

This explicit type casting is a good fix for the dtype mismatch error in the CI.

As you mentioned in the PR description, this is a temporary workaround. To improve long-term maintainability and ensure this is addressed later, I suggest adding a TODO comment here. This will make the reason for the cast clear to anyone reading the code in the future.

Suggested change
gating_output.to(dtype=torch.float32),
# TODO: Cast to float32 to match correction_bias.dtype. This is a temporary
# workaround for a limitation in the aiter kernel. Remove this cast once
# bf16/bf16 mixed-precision GEMM is supported.
# Ref: https://github.com/sgl-project/sglang/pull/7825
gating_output.to(dtype=torch.float32),

correction_bias,
topk_weights,
topk_ids,
Expand Down
Loading