move check to base class

Signed-off-by: Chendi Xue <chendi.xue@intel.com>

Author: xuechendi

vllm_omni/platforms/cuda/platform.py

@@ -115,7 +115,3 @@ def get_free_memory(cls, device: torch.device | None = None) -> int:
     @classmethod
     def get_device_name(cls, device_id: int = 0) -> str:
         return torch.cuda.get_device_name(device_id)
-
-    @classmethod
-    def supports_cpu_offload(cls) -> bool:
-        return True

vllm_omni/platforms/interface.py

@@ -98,6 +98,10 @@ def synchronize(cls) -> None:
     def get_free_memory(cls, device: torch.device | None = None) -> int:
         raise NotImplementedError
 
+    @classmethod
+    def supports_cpu_offload(cls) -> bool:
+        return True
+
 
 class UnspecifiedOmniPlatform(OmniPlatform):
     _omni_enum = OmniPlatformEnum.UNSPECIFIED

vllm_omni/platforms/npu/platform.py

@@ -86,7 +86,3 @@ def get_free_memory(cls, device: torch.device | None = None) -> int:
     def get_device_total_memory(cls, device_id: int = 0) -> int:
         device_props = torch.npu.get_device_properties(device_id)
         return device_props.total_memory
-
-    @classmethod
-    def supports_cpu_offload(cls) -> bool:
-        return False

vllm_omni/platforms/rocm/platform.py

@@ -99,7 +99,3 @@ def synchronize(cls) -> None:
     def get_free_memory(cls, device: torch.device | None = None) -> int:
         free, _ = torch.cuda.mem_get_info(device)
         return free
-
-    @classmethod
-    def supports_cpu_offload(cls) -> bool:
-        return True

vllm_omni/platforms/xpu/platform.py

@@ -79,7 +79,3 @@ def get_free_memory(cls, device: torch.device | None = None) -> int:
             device_id = device.index if device.index is not None else 0
         props = torch.xpu.get_device_properties(device_id)
         return props.total_memory
-
-    @classmethod
-    def supports_cpu_offload(cls) -> bool:
-        return True