xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-04-11 03:47:03 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] Backport request id validation to v0 (#11036)

Browse Source 
Signed-off-by: Joe Runde <Joseph.Runde@ibm.com>
This commit is contained in:
Joe Runde 2024-12-10 09:38:23 -07:00 committed by GitHub
parent d05f88679b
commit 9b9cef3145
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/engine/multiprocessing/client.py
View File

@ -576,6 +576,10 @@ class MQLLMEngineClient(EngineClient):
if self._errored_with is not None:
raise ENGINE_DEAD_ERROR(self._errored_with)
# Ensure the request id is unique among running requests
if request_id in self.output_queues:
raise ValueError(f"Request {request_id} already exists")
# Constructing guided decoding logits processors is expensive, so we do
# it here to avoid contending with cpu resources and the GIL on the
# backend process.

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

@ -152,7 +152,7 @@ class AsyncLLM(EngineClient):
"""Add new request to the AsyncLLM."""
if self.detokenizer.is_request_active(request_id):
raise KeyError(f"Request {request_id} already exists.")
raise ValueError(f"Request {request_id} already exists.")
# 1) Create a new AsyncStream for the request.
stream = self._add_request_to_streams(request_id)