Skip to content

[Generative Score API] Fix on prefill-only scheduler running batch loss track problem #14320

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

Open
haNa-meister wants to merge 4 commits into
base: main
Choose a base branch
from
Open

[Generative Score API] Fix on prefill-only scheduler running batch loss track problem #14320

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
c6e4086
debug: fix running batch tracking for prefill-only batch
haNa-meister Dec 2, 2025
8794861
fix filter running batch
haNa-meister Dec 2, 2025
5eb5c20
format
haNa-meister Dec 2, 2025
68b1b97
Merge branch 'main' into main
sundar24295s Feb 4, 2026
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
12 changes: 9 additions & 3 deletions python/sglang/srt/managers/scheduler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1841,14 +1841,17 @@ def get_next_batch_to_run(self) -> Optional[ScheduleBatch]:
self.running_batch.batch_is_full = False

# Merge the new batch into the running batch.
# For prefill-only batch, we can avoid going through decoding step.
if not self.last_batch.is_empty() and not self.last_batch.is_prefill_only:
if not self.last_batch.is_empty():
if self.running_batch.is_empty():
self.running_batch = self.last_batch
else:
# Merge running_batch with prefill batch
self.running_batch.merge_batch(self.last_batch)

# For prefill-only running batch, filter out finished requests since it won't run decode.
if self.running_batch.is_prefill_only:
self.running_batch.filter_batch()

new_batch = self.get_new_batch_prefill()

need_mlp_sync = self.require_mlp_sync
Expand All @@ -1867,7 +1870,10 @@ def get_next_batch_to_run(self) -> Optional[ScheduleBatch]:
ret = new_batch
else:
# Run decode
if not self.running_batch.is_empty():
if (
not self.running_batch.is_empty()
and not self.running_batch.is_prefill_only
):
self.running_batch = self.update_running_batch(self.running_batch)
ret = self.running_batch if not self.running_batch.is_empty() else None
else:
Expand Down
Loading