A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 22:35 UTC
[XPU] use `ZE_AFFINITY_MASK` for device select on xpu (#21815)
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
Kunshang Ji committed
Jul 30, 2025 at 11:56 UTC
05cbbe20c55d957f18c12f7eb11cf551504e657d
1 file changed
+1
-1
vllm/platforms/xpu.py
+1
-1
@@ -30,7 +30,7 @@ class XPUPlatform(Platform):
30
# see https://github.com/ray-project/ray/blob/6a5eb5865eeb9ccf058a79b44f107e327e360673/python/ray/_private/accelerators/intel_gpu.py#L20 # noqa: E501
31
ray_device_key: str = "GPU"
32
dist_backend: str = "ccl" # ccl | xccl
33
- device_control_env_var: str = "ONEAPI_DEVICE_SELECTOR"
33
+ device_control_env_var: str = "ZE_AFFINITY_MASK"
34
35
@classmethod
36
def get_attn_backend_cls(cls, selected_backend: _Backend, head_size: int,