biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Platform] Allow platform use V1 Engine by default (#19792)

Browse Source 
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
wangxiyuan
2025-06-18 21:03:36 +08:00
committed by GitHub
parent cca91a7a10
commit 257ab95439
3 changed files with 20 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
vllm/engine/arg_utils.py
View File

@@ -1303,7 +1303,7 @@ class EngineArgs:
# Skip this check if we are running on a non-GPU platform,
# or if the device capability is not available
# (e.g. in a Ray actor without GPUs).
from vllm.platforms import CpuArchEnum, current_platform
from vllm.platforms import current_platform
if (current_platform.is_cuda()
and current_platform.get_device_capability()
and current_platform.get_device_capability().major < 8):
@@ -1445,14 +1445,10 @@ class EngineArgs:
_raise_or_fallback(feature_name=name, recommend_to_remove=False)
return False
# Non-[CUDA, TPU, x86 CPU] may be supported on V1,
# but off by default for now.
v0_hardware = not any(
(current_platform.is_cuda_alike(), current_platform.is_tpu(),
(current_platform.is_cpu()
and current_platform.get_cpu_architecture() == CpuArchEnum.X86)))
if v0_hardware and _warn_or_fallback( # noqa: SIM103
current_platform.device_name):
# The platform may be supported on V1, but off by default for now.
if not current_platform.default_v1( # noqa: SIM103
model_config=model_config) and _warn_or_fallback(
current_platform.device_name):
return False
#############################################################