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 15, 03:30 UTC
[XPU] log clean up for XPU platform (#20553)
Signed-off-by: yan <yan.ma@intel.com>
Yan Ma committed
Jul 7, 2025 at 16:38 UTC
3112271f6e5d50b3d94a2efa88a5a8e77826b897
2 files changed
+4
-4
vllm/_custom_ops.py
+2
-1
@@ -13,7 +13,8 @@ from vllm.scalar_type import ScalarType
13
14
logger = init_logger(__name__)
15
16
-if not current_platform.is_tpu() and not current_platform.is_hpu():
16
+if not current_platform.is_tpu() and not current_platform.is_hpu()\
17
+ and not current_platform.is_xpu():
18
try:
19
import vllm._C
20
except ImportError as e:
vllm/platforms/xpu.py
+2
-3
@@ -37,7 +37,7 @@ class XPUPlatform(Platform):
37
dtype: torch.dtype, kv_cache_dtype: Optional[str],
38
block_size: int, use_v1: bool,
39
use_mla: bool) -> str:
40
- if selected_backend != _Backend.IPEX:
40
+ if selected_backend is not None and selected_backend != _Backend.IPEX:
41
logger.info("Cannot use %s backend on XPU.", selected_backend)
42
use_v1 = envs.VLLM_USE_V1
43
if not use_v1:
@@ -133,8 +133,7 @@ class XPUPlatform(Platform):
133
134
@classmethod
135
def is_pin_memory_available(cls):
136
- logger.warning("Pin memory is not supported on XPU.")
137
- return False
136
+ return True
137
138
@classmethod
139
def get_current_memory_usage(cls,