[Bugfix] Fix aclgraph not enabled by default (#2590)

### What this PR does / why we need it?
As vllm will set `cudagraph_mode` to `NONE` before
`check_and_update_config` in post init of `VllmConfig`
(5da4f5d857/vllm/config/__init__.py (L3630)),
we always have `cudagraph_mode` isn't `None`, thus we must remove this
check and add it when the related adaption in vllm is done.

part of https://github.com/vllm-project/vllm-ascend/pull/2577, will add
the e2e test on applying reply after the CI refactor is done

### How was this patch tested?
CI passed with existing test.

- vLLM version: v0.10.1.1
- vLLM main:
f48a9af892

Signed-off-by: MengqingCao <cmq0113@163.com>
This commit is contained in:
Mengqing Cao
2025-08-28 14:08:31 +08:00
committed by GitHub
parent cf96366a39
commit 6c973361fc
3 changed files with 22 additions and 20 deletions

View File

@@ -13,12 +13,10 @@ from vllm.compilation.cuda_graph import CUDAGraphOptions
from vllm.compilation.monitor import validate_cudagraph_capturing_enabled
from vllm.config import CUDAGraphMode, VllmConfig
from vllm.forward_context import BatchDescriptor, get_forward_context
from vllm.logger import init_logger
from vllm.logger import logger
from vllm.platforms import current_platform
from vllm.utils import weak_ref_tensors
logger = init_logger(__name__)
@dataclasses.dataclass
class ACLGraphEntry:
@@ -182,5 +180,6 @@ class ACLGraphWrapper:
f"during replay. Expected {entry.input_addresses}, "
f"got {new_input_addresses}")
logger.info_once("Replaying aclgraph")
entry.aclgraph.replay()
return entry.output