xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-19 07:55:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Refactor][MISC] del redundant code in ParallelConfig.postinit (#10614)

Browse Source 
Signed-off-by: MengqingCao <cmq0113@163.com>
This commit is contained in:
Mengqing Cao 2024-11-25 13:14:56 +08:00 committed by GitHub
parent 214efc2c3c
commit 7ea3cd7c3e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 5 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
vllm/config.py
View File

@ -998,20 +998,15 @@ class ParallelConfig:
raise ValueError(f"worker-use-ray can't be used with " raise ValueError(f"worker-use-ray can't be used with "
f"distributed executor backend " f"distributed executor backend "
f"'{self.distributed_executor_backend}'.") f"'{self.distributed_executor_backend}'.")
ray_only_devices = ["tpu", "hpu"]
if current_platform.is_tpu() and self.world_size > 1: if (current_platform.device_type in ray_only_devices
and self.world_size > 1):
if self.distributed_executor_backend is None: if self.distributed_executor_backend is None:
self.distributed_executor_backend = "ray" self.distributed_executor_backend = "ray"
if self.distributed_executor_backend != "ray": if self.distributed_executor_backend != "ray":
raise ValueError( raise ValueError(
"TPU backend only supports Ray for distributed inference.") f"{current_platform.device_type.upper()} backend only "
"supports Ray for distributed inference.")
if current_platform.is_hpu() and self.world_size > 1:
if self.distributed_executor_backend is None:
self.distributed_executor_backend = "ray"
if self.distributed_executor_backend != "ray":
raise ValueError(
"HPU backend only supports Ray for distributed inference.")
if self.distributed_executor_backend is None and self.world_size > 1: if self.distributed_executor_backend is None and self.world_size > 1:
# We use multiprocessing by default if world_size fits on the # We use multiprocessing by default if world_size fits on the