xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 03:05:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

[XPU] Fix circular import error. (#24927)

Browse Source 
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
This commit is contained in:
Kunshang Ji 2025-09-16 11:35:36 +08:00 committed by GitHub
parent 0af3ce1355
commit 5206ab20ba
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/platforms/xpu.py
View File

@ -9,7 +9,6 @@ import torch
import vllm.envs as envs
from vllm.logger import init_logger
from vllm.utils import DEFAULT_MAX_NUM_BATCHED_TOKENS
from vllm.v1.attention.backends.utils import set_kv_cache_layout
from .interface import DeviceCapability, Platform, PlatformEnum, _Backend
@ -164,11 +163,16 @@ class XPUPlatform(Platform):
vllm_config.scheduler_config.max_num_batched_tokens = max(
vllm_config.scheduler_config.max_model_len,
DEFAULT_MAX_NUM_BATCHED_TOKENS)
from vllm.v1.attention.backends.utils import set_kv_cache_layout
set_kv_cache_layout("NHD")
logger.info("Setting VLLM_KV_CACHE_LAYOUT to 'NHD' for XPU; "
"only NHD layout is supported by XPU attention kernels.")
@classmethod
def support_hybrid_kv_cache(cls) -> bool:
return True
@classmethod
def is_pin_memory_available(cls):
return True