[Bugfix] Fix SHM cache initialization (#26427)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-10-09 02:48:04 -07:00
committed by GitHub
parent dc7976dd9f
commit 4bdf7ac593
30 changed files with 357 additions and 417 deletions
@@ -10,7 +10,6 @@ from typing import Optional, Union, cast
import jinja2
from fastapi import Request
from vllm.config import ModelConfig
from vllm.engine.protocol import EngineClient
from vllm.entrypoints.logger import RequestLogger
from vllm.entrypoints.openai.protocol import (
@@ -44,7 +43,6 @@ class OpenAIServingCompletion(OpenAIServing):
def __init__(
self,
engine_client: EngineClient,
model_config: ModelConfig,
models: OpenAIServingModels,
*,
request_logger: Optional[RequestLogger],
@@ -55,7 +53,6 @@ class OpenAIServingCompletion(OpenAIServing):
):
super().__init__(
engine_client=engine_client,
model_config=model_config,
models=models,
request_logger=request_logger,
return_tokens_as_token_ids=return_tokens_as_token_ids,
@@ -201,7 +198,7 @@ class OpenAIServingCompletion(OpenAIServing):
# but pre-commit in CI fails without it.
engine_prompt = cast(Union[EmbedsPrompt, TokensPrompt], engine_prompt)
if isinstance(sampling_params, BeamSearchParams):
generator = self.engine_client.beam_search(
generator = self.beam_search(
prompt=engine_prompt,
request_id=request_id,
params=sampling_params,