[Kernel][SM100] Add a CuTeDSL fused query kernel (#49792)
Signed-off-by: Peiyuan Zhou <peiyuanzhou1994@gmail.com> Co-authored-by: Claude <noreply@anthropic.com>
This commit is contained in:
@@ -19,6 +19,7 @@ from cutlass.cutlass_dsl import T, dsl_user_op
|
||||
_TORCH_TO_CUTE_DTYPE = {
|
||||
torch.bfloat16: BFloat16,
|
||||
torch.float8_e4m3fn: Float8E4M3FN,
|
||||
torch.float32: Float32,
|
||||
}
|
||||
|
||||
_CUTE_TO_PTX_DTYPE = {
|
||||
|
||||
Reference in New Issue
Block a user