xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-07-10 00:27:23 +08:00
Code Issues Packages Projects Releases Wiki Activity

[XPU] Remove distributed_executor_backend check (#30760)

Browse Source 
Signed-off-by: sihao.li <sihao.li@intel.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
This commit is contained in:
sihao_li 2025-12-24 13:34:33 +08:00 committed by GitHub
parent bb24592d13
commit 471ddb99a0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 0 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
vllm/platforms/xpu.py
View File

@ -7,7 +7,6 @@ from typing import TYPE_CHECKING, Optional
import torch import torch
import vllm.envs as envs
from vllm.attention.backends.registry import AttentionBackendEnum from vllm.attention.backends.registry import AttentionBackendEnum
from vllm.logger import init_logger from vllm.logger import init_logger
@ -168,32 +167,6 @@ class XPUPlatform(Platform):
if vllm_config.kv_transfer_config is not None: if vllm_config.kv_transfer_config is not None:
vllm_config.kv_transfer_config.enable_permute_local_kv = True vllm_config.kv_transfer_config.enable_permute_local_kv = True
if parallel_config.distributed_executor_backend is None:
if parallel_config.world_size > 1:
parallel_config.distributed_executor_backend = "ray"
else:
parallel_config.distributed_executor_backend = "uni"
elif parallel_config.distributed_executor_backend == "mp":
# FIXME(kunshang):
# spawn needs calling `if __name__ == '__main__':`
# fork is not supported for xpu start new process.
if envs.VLLM_WORKER_MULTIPROC_METHOD != "spawn":
os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
logger.warning(
"Please use spawn as start method if you want to use mp."
)
elif (
parallel_config.distributed_executor_backend != "ray"
and parallel_config.distributed_executor_backend != "uni"
and parallel_config.distributed_executor_backend != "external_launcher"
):
logger.warning(
"%s is not supported on XPU, fallback to ray distributed"
" executor backend.",
parallel_config.distributed_executor_backend,
)
parallel_config.distributed_executor_backend = "ray"
if model_config and model_config.use_mla: if model_config and model_config.use_mla:
logger.info( logger.info(
"MLA is enabled on a non-GPU platform; forcing chunked " "MLA is enabled on a non-GPU platform; forcing chunked "