[XPU][NPU][ROCM] enable cpu_offloading flag for non_cuda (#1488)

Signed-off-by: gcanlin <canlinguosdu@gmail.com>
Signed-off-by: Chendi Xue <chendi.xue@intel.com>
Co-authored-by: gcanlin <canlinguosdu@gmail.com>
This commit is contained in:
Chendi.Xueandgcanlin authored and GitHub committed 2026-02-27 14:52:15 +08:00
1 parent ba9594724e
commit 482f9b89a1
6 files changed
+16 -11

No files matched your search

+4
View File
@@ -98,6 +98,10 @@ class OmniPlatform(Platform):
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