Skip to content

Commit 1f0f228

Browse files
committed
[Fix] Each request in the decode instance encounters a load failure
1 parent 99b09be commit 1f0f228

File tree

1 file changed

+15
-6
lines changed

1 file changed

+15
-6
lines changed

ucm/integration/vllm/uc_connector.py

Lines changed: 15 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -492,14 +492,23 @@ def get_finished(self, finished_req_ids: set[str]) -> tuple[set[str], set[str]]:
492492
continue
493493
unfinished_tasks = []
494494
for task in tasks:
495-
ret = self.connector.check(task)
496-
if ret == -1:
495+
ret, finish = self.connector.check(task)
496+
if ret != 0:
497+
logger.error(
498+
f"Task {task} failed, check return {ret} for request {req_id}"
499+
)
500+
self._load_failed_reqs.add(req_id)
501+
break
502+
if not finish:
497503
unfinished_tasks.append(task)
498504
continue
499-
elif ret == 0 and self.connector.wait(task) == 0:
500-
continue
501-
self._load_failed_reqs.add(req_id)
502-
break
505+
wret = self.connector.wait(task)
506+
if wret != 0:
507+
logger.error(
508+
f"Task {task} failed, wait return {wret} for request {req_id}"
509+
)
510+
self._load_failed_reqs.add(req_id)
511+
break
503512
if not unfinished_tasks:
504513
done_recving.add(req_id)
505514
self._need_load_reqs[req_id] = unfinished_tasks

0 commit comments

Comments
 (0)