[V0 Deprecation] Remove AsyncLLMEngine (#25025)

Signed-off-by: Woosuk Kwon <woosuk@thinkingmachines.ai>
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
Woosuk Kwon
2025-09-18 11:07:42 -07:00
committed by GitHub
parent 505805b645
commit e19bce40a1
11 changed files with 76 additions and 1967 deletions
-18
View File
@@ -7,7 +7,6 @@ import pytest
import vllm.envs as envs
from vllm import LLM
from vllm.engine.arg_utils import AsyncEngineArgs
from vllm.engine.async_llm_engine import AsyncLLMEngine
MODEL = "meta-llama/Llama-3.2-1B-Instruct"
@@ -96,20 +95,3 @@ def test_v1_attn_backend(monkeypatch):
_ = AsyncEngineArgs(model=MODEL).create_engine_config()
assert envs.VLLM_USE_V1
m.delenv("VLLM_USE_V1")
def test_reject_using_constructor_directly(monkeypatch):
with monkeypatch.context() as m:
if os.getenv("VLLM_USE_V1", None):
m.delenv("VLLM_USE_V1")
# Sets VLLM_USE_V1=1.
vllm_config = AsyncEngineArgs(model=MODEL).create_engine_config()
# This uses the V0 constructor directly.
with pytest.raises(ValueError):
AsyncLLMEngine(vllm_config,
AsyncLLMEngine._get_executor_cls(vllm_config),
log_stats=True)
m.delenv("VLLM_USE_V1")