Skip to content

fix layer_norm grad #78567

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
zhengshengning merged 1 commit into from
Apr 3, 2026
Merged

fix layer_norm grad #78567

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions paddle/phi/kernels/gpu/layer_norm_grad_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,6 +39,9 @@ static inline LayerNormGadKernelVariant LayerNormGradKernelDispatch(
const DenseTensor* scale,
const DenseTensor* bias) {
#if defined(PADDLE_WITH_CUDA) && !defined(PADDLE_WITH_HIP) && !defined(_WIN32)
if (FLAGS_use_accuracy_compatible_kernel) {
return LayerNormGadKernelVariant::GENERIC;
}
if (scale != nullptr && bias != nullptr &&
input_type != paddle::DataType::FLOAT32 && hidden_size != 4096 &&
hidden_size > 1024 && hidden_size <= 10240 &&
Expand Down
40 changes: 3 additions & 37 deletions paddle/phi/kernels/gpu/rms_norm_cuda_kernel.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -731,37 +731,6 @@ void LayerNormFwdCompatKernel(

auto stream = dev_ctx.stream();

// if (!FLAGS_use_accuracy_compatible_kernel && rows <= 1024 &&
// (cols / rows >= 32)) {
// constexpr int num_vec_elems2 = 8;
// constexpr int alignment2 = num_vec_elems2 * sizeof(T);
// bool can_vec_X2 = can_vectorize(x_data, alignment2);
// bool can_vec_Y2 = can_vectorize(y_data, alignment2);
// bool can_vec_gamma2 = can_vectorize(gamma_data, alignment2);
// bool can_vec_beta2 = can_vectorize(beta_data, alignment2);
// bool is_supported_type2 = (std::is_same<T, phi::dtype::float16>::value ||
// std::is_same<T, phi::dtype::bfloat16>::value);
// if (is_supported_type2 &&
// cols <=
// static_cast<int64_t>(1ULL << std::numeric_limits<float>::digits)
// &&
// cols % num_vec_elems2 == 0 && can_vec_X2 && can_vec_Y2 &&
// can_vec_gamma2 && can_vec_beta2) {
// launch_vectorized_layer_norm_kernel_driver<T, T_ACC, 8>(
// cols,
// rows,
// static_cast<T_ACC>(epsilon),
// x_data,
// gamma_data,
// beta_data,
// y_data,
// mean_data,
// var_data,
// stream);
// return;
// }
// }

// Check vectorization conditions for vec_size=4
constexpr int num_vec_elems = 4;
constexpr int alignment = num_vec_elems * sizeof(T);
Expand Down Expand Up @@ -1555,6 +1524,7 @@ __device__ __inline__ void layer_norm_compute_gI(const T* __restrict__ dY,
}

stats_x1 = BlockReduceSum(stats_x1, buf);
__syncthreads();
stats_x2 = BlockReduceSum(stats_x2, buf);
if (threadIdx.x == 0) {
buf[0] = stats_x1;
Expand Down Expand Up @@ -1658,6 +1628,7 @@ __global__ void layer_norm_grad_input_kernel_vectorized(

// Reduction in Shared Memory
stats_x1 = BlockReduceSum(stats_x1, reduce_buf);
__syncthreads();
stats_x2 = BlockReduceSum(stats_x2, reduce_buf);
if (threadIdx.x == 0) {
reduce_buf[0] = stats_x1;
Expand Down Expand Up @@ -2084,12 +2055,7 @@ void LayerNormBwdCompatKernel(
constexpr int num_threads = 128;
constexpr int nshared = (num_threads / kWarpSize) * sizeof(T_ACC);

if (!FLAGS_use_accuracy_compatible_kernel && is_supported_type2 &&
bAlignedBuffers2 && (N % 8 == 0 && M <= 1024 && (N / M >= 32))) {
layer_norm_grad_input_kernel_vectorized<T, T_ACC, 8>
<<<blocks, num_threads, nshared, stream>>>(
dY_data, X_data, mean_data, rstd_data, gamma_data, dX_data, N);
} else if (is_supported_type && bAlignedBuffers && bVectorSizeMultiple) {
if (is_supported_type && bAlignedBuffers && bVectorSizeMultiple) {
layer_norm_grad_input_kernel_vectorized<T, T_ACC, kVecSize>
<<<blocks, num_threads, nshared, stream>>>(
dY_data, X_data, mean_data, rstd_data, gamma_data, dX_data, N);
Expand Down
Loading