xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-03 02:44:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bug] [V1] Try fetching stop_reason from EngineOutput before checking the request (#13108)

Browse Source
This commit is contained in:
bnellnm 2025-02-12 05:39:16 -05:00 committed by GitHub
parent f1042e86f0
commit f4d97e4fc2
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 6 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
vllm/v1/engine/output_processor.py
View File

@ -2,7 +2,7 @@
import asyncio import asyncio
from dataclasses import dataclass from dataclasses import dataclass
from typing import Dict, List, Optional from typing import Dict, List, Optional, Union
from vllm.outputs import RequestOutput from vllm.outputs import RequestOutput
from vllm.sampling_params import RequestOutputKind from vllm.sampling_params import RequestOutputKind
@ -164,6 +164,7 @@ class OutputProcessor:
new_token_ids = engine_core_output.new_token_ids new_token_ids = engine_core_output.new_token_ids
finish_reason = engine_core_output.finish_reason finish_reason = engine_core_output.finish_reason
stop_reason = engine_core_output.stop_reason
# TODO(andy): prompt logprobs + chunked prefill can # TODO(andy): prompt logprobs + chunked prefill can
# result in engine core returning an output for a # result in engine core returning an output for a
@ -181,9 +182,10 @@ class OutputProcessor:
# 2) Detokenize the token ids into text and check for stop # 2) Detokenize the token ids into text and check for stop
# strings. # strings.
stop_reason = req_state.detokenizer.update(new_token_ids) stop_string = req_state.detokenizer.update(new_token_ids)
if stop_reason: if stop_string and finish_reason != FinishReason.STOP:
finish_reason = FinishReason.STOP finish_reason = FinishReason.STOP
stop_reason = stop_string
# 3) Compute sample and prompt logprobs for request, # 3) Compute sample and prompt logprobs for request,
# if required. # if required.
@ -250,7 +252,7 @@ class OutputProcessor:
request_state: RequestState, request_state: RequestState,
new_token_ids: List[int], new_token_ids: List[int],
finish_reason: Optional[FinishReason], finish_reason: Optional[FinishReason],
stop_reason: Optional[str], stop_reason: Union[int, str, None],
) -> Optional[RequestOutput]: ) -> Optional[RequestOutput]:
finished = finish_reason is not None finished = finish_reason is not None