[XPU] Route weightless RMSNorm to _C dispatch (#47121)
Signed-off-by: Yintong Lu <yintong.lu@intel.com> Co-authored-by: Yongqi Wang <yongqi.wang@intel.com> Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
This commit is contained in:
co-authored by
Yongqi Wang
Kunshang Ji
parent
f51193b9ae
commit
72297d859b
@@ -30,9 +30,6 @@ def rms_norm(
|
|||||||
x: Tensor, weight: Tensor | None, epsilon: float, variance_size: int | None = None
|
x: Tensor, weight: Tensor | None, epsilon: float, variance_size: int | None = None
|
||||||
) -> Tensor:
|
) -> Tensor:
|
||||||
assert variance_size is None
|
assert variance_size is None
|
||||||
if weight is None:
|
|
||||||
# Kernel requires weight tensor, pass ones
|
|
||||||
weight = torch.ones(x.shape[-1], device=x.device, dtype=x.dtype)
|
|
||||||
output = torch.empty(x.shape, device=x.device, dtype=x.dtype)
|
output = torch.empty(x.shape, device=x.device, dtype=x.dtype)
|
||||||
torch.ops._C.rms_norm(output, x, weight, epsilon)
|
torch.ops._C.rms_norm(output, x, weight, epsilon)
|
||||||
return output
|
return output
|
||||||
@@ -58,8 +55,5 @@ def fused_add_rms_norm(
|
|||||||
variance_size: int | None = None,
|
variance_size: int | None = None,
|
||||||
) -> tuple[Tensor, Tensor]:
|
) -> tuple[Tensor, Tensor]:
|
||||||
assert variance_size is None
|
assert variance_size is None
|
||||||
if weight is None:
|
|
||||||
# Kernel requires weight tensor, pass ones
|
|
||||||
weight = torch.ones(x.shape[-1], device=x.device, dtype=x.dtype)
|
|
||||||
torch.ops._C.fused_add_rms_norm(x, x_residual, weight, epsilon)
|
torch.ops._C.fused_add_rms_norm(x, x_residual, weight, epsilon)
|
||||||
return x, x_residual
|
return x, x_residual
|
||||||
|
|||||||
Reference in New Issue
Block a user