xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-09 12:25:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

updated

Browse Source 
Signed-off-by: Robert Shaw <rshaw@neuralmagic.com>
This commit is contained in:
Robert Shaw 2025-10-03 13:47:16 -04:00
parent 52a7d91980
commit 6f62c94d7e
2 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/v1/engine/core.py
View File

@ -1050,6 +1050,7 @@ class DPEngineCoreProc(EngineCoreProc):
self._maybe_publish_request_counts()
local_unfinished_reqs = self.scheduler.has_unfinished_requests()
logger.info(f"{local_unfinished_reqs=}")
if not executed:
if not local_unfinished_reqs and not self.engines_running:
# All engines are idle.
@ -1057,6 +1058,7 @@ class DPEngineCoreProc(EngineCoreProc):
# We are in a running state and so must execute a dummy pass
# if the model didn't execute any ready requests.
logger.info("===EXECUTE_DUMMY_BATCH===")
self.execute_dummy_batch()
# 3) All-reduce operation to determine global unfinished reqs.

2
vllm/v1/engine/core_client.py
View File

@ -1077,7 +1077,7 @@ class DPAsyncMPClient(AsyncMPClient):
if counts is not None:
sliced_counts = counts[count_slice]
self.lb_engines = sliced_counts
logger.debug("Received counts: %s (%s)", sliced_counts,
logger.debug("Received counts: %s (%s)", counts,
count_slice)
resources.stats_update_task = asyncio.create_task(