Skip to content

[Bugfix] Fix error with penalties when speculative decoding and structural output are enabled #26586

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 3 commits into from
Oct 19, 2025
Merged

[Bugfix] Fix error with penalties when speculative decoding and structural output are enabled #26586

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
e773ad5
Fix error with penalties when speculative decoding and structural out…
southfreebird Oct 10, 2025
e4cc330
Fix comments + pre-commit
southfreebird Oct 10, 2025
1c3d127
Merge branch 'vllm-project:main' into fix/logit-processors-with-spec-…
southfreebird Oct 19, 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
10 changes: 8 additions & 2 deletions vllm/v1/worker/gpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -780,7 +780,7 @@ def _update_states(self, scheduler_output: "SchedulerOutput") -> None:

# Add spec_token_ids to token_ids_cpu.
spec_token_ids = scheduler_output.scheduled_spec_decode_tokens.get(
req_id, ()
req_id, []
)
if spec_token_ids:
num_spec_tokens = len(spec_token_ids)
Expand All @@ -791,7 +791,13 @@ def _update_states(self, scheduler_output: "SchedulerOutput") -> None:
] = spec_token_ids
# NOTE(woosuk): `num_tokens` here may include spec tokens.
self.input_batch.num_tokens[req_index] += num_spec_tokens
self.input_batch.spec_token_ids[req_index] = spec_token_ids

# When speculative decoding is used with structured output,
# the scheduler can drop draft tokens that do not
# conform to the schema. This can result in
# scheduler_output.scheduled_spec_decode_tokens being empty,
# even when speculative decoding is enabled.
self.input_batch.spec_token_ids[req_index] = spec_token_ids

# Add the new or resumed requests to the persistent batch.
# The smaller empty indices are filled first.
Expand Down