fix arm test_norm.py error (#30597)
This commit is contained in:
@@ -397,6 +397,7 @@ void norm4d_kernel_impl(
|
|||||||
const scalar_t* __restrict__ input,
|
const scalar_t* __restrict__ input,
|
||||||
const NormParams& p,
|
const NormParams& p,
|
||||||
const scalar_t* __restrict__ gate = nullptr) {
|
const scalar_t* __restrict__ gate = nullptr) {
|
||||||
|
#if defined(CPU_CAPABILITY_AVX512)
|
||||||
// fast path only applies to bfloat16 when D in {32, 64, 128, 256, 512}
|
// fast path only applies to bfloat16 when D in {32, 64, 128, 256, 512}
|
||||||
if constexpr (std::is_same_v<scalar_t, at::BFloat16>) {
|
if constexpr (std::is_same_v<scalar_t, at::BFloat16>) {
|
||||||
switch (p.D) {
|
switch (p.D) {
|
||||||
@@ -409,6 +410,7 @@ void norm4d_kernel_impl(
|
|||||||
break;
|
break;
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
#endif
|
||||||
|
|
||||||
// generic path
|
// generic path
|
||||||
LAUNCH_PARALLEL_LOOP(
|
LAUNCH_PARALLEL_LOOP(
|
||||||
|
|||||||
Reference in New Issue
Block a user