[AMD][Fix] Fix regression in jit build error with FLUX.2-dev on gfx1250 (#38329)
This commit is contained in:
@@ -15,8 +15,13 @@ _ALIGNMENT = 32
|
|||||||
|
|
||||||
|
|
||||||
def _blackwell_or_newer(device: torch.device) -> bool:
|
def _blackwell_or_newer(device: torch.device) -> bool:
|
||||||
|
# ROCm reports a capability derived from the gfx number, so gfx1250 comes
|
||||||
|
# back as (12, 5) and clears a bare major >= 10 test. The kernel is PTX, so
|
||||||
|
# require a CUDA build like rope/qwen_qkv_epilogue_jit.py does.
|
||||||
return (
|
return (
|
||||||
torch.cuda.is_available() and torch.cuda.get_device_capability(device)[0] >= 10
|
torch.version.cuda is not None
|
||||||
|
and torch.cuda.is_available()
|
||||||
|
and torch.cuda.get_device_capability(device)[0] >= 10
|
||||||
)
|
)
|
||||||
|
|
||||||
|
|
||||||
|
|||||||
Reference in New Issue
Block a user