Skip to content

[Bugfix] BailingMoeV3 KDA: skip absent metadata during CUDA graph profiling #53593

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
njhill merged 1 commit into from
Aug 25, 2026
Merged

[Bugfix] BailingMoeV3 KDA: skip absent metadata during CUDA graph profiling #53593

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
5 changes: 4 additions & 1 deletion vllm/model_executor/models/bailing_moe_v3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -813,7 +813,10 @@ def _forward(
return

assert isinstance(attn_metadata_map, dict)
attn_metadata = attn_metadata_map[self.prefix]
attn_metadata = attn_metadata_map.get(self.prefix)
if attn_metadata is None:
# Profile/warmup dummy runs skip mamba-family metadata.
return
assert isinstance(attn_metadata, GDNAttentionMetadata)
has_initial_state = attn_metadata.has_initial_state
spec_query_start_loc = attn_metadata.spec_query_start_loc
Expand Down
Loading