xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-01-01 07:27:28 +08:00
Code Issues Packages Projects Releases Wiki Activity

[V1][Metrics] Fix traceback with preemptions+LoRA (#14220)

Browse Source 
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
This commit is contained in:
Mark McLoughlin 2025-03-07 20:36:16 +00:00 committed by GitHub
parent 8ed5421aaa
commit e1f0835ae0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 8 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/v1/metrics/stats.py
View File

@ -150,6 +150,7 @@ class IterationStats:
LoRARequestStates.scheduled_request(lora_stats, req_id)
elif event.type == EngineCoreEventType.PREEMPTED:
self.num_preempted_reqs += 1
LoRARequestStates.preempted_request(lora_stats, req_id)
def update_from_finished_request(self, finish_reason: "FinishReason",
num_prompt_tokens: int,
@ -224,6 +225,13 @@ class LoRARequestStates:
lora_stats.waiting_requests.remove(request_id)
lora_stats.running_requests.add(request_id)
@staticmethod
def preempted_request(lora_stats: Optional[LoRAStats], request_id: str):
if lora_stats is None:
return
lora_stats.running_requests.remove(request_id)
lora_stats.waiting_requests.add(request_id)
def update_iteration_stats(self,
iteration_stats: Optional[IterationStats]):
if iteration_stats is None: