add qwen3

vllm-v0.6.2/vllm/platforms/xpu.py

@@ -0,0 +1,26 @@
+import torch
+
+from .interface import DeviceCapability, Platform, PlatformEnum
+
+
+class XPUPlatform(Platform):
+    _enum = PlatformEnum.XPU
+
+    @staticmethod
+    def get_device_capability(device_id: int = 0) -> DeviceCapability:
+        major, minor, *_ = torch.xpu.get_device_capability(
+            device_id)['version'].split('.')
+        return DeviceCapability(major=int(major), minor=int(minor))
+
+    @staticmethod
+    def get_device_name(device_id: int = 0) -> str:
+        return torch.xpu.get_device_name(device_id)
+
+    @classmethod
+    def get_device_total_memory(cls, device_id: int = 0) -> int:
+        device_props = torch.xpu.get_device_properties(device_id)
+        return device_props.total_memory
+
+    @staticmethod
+    def inference_mode():
+        return torch.no_grad()