Skip to content

Align matmul grad with torch v2.9.1 #76922

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

Open
A-nnonymous wants to merge 2 commits into
base: develop
Choose a base branch
from
Open

Align matmul grad with torch v2.9.1 #76922

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
206f82c
combine batch with aggregate dim, save reduce, align with torch
A-nnonymous Dec 15, 2025
a1b3234
fix miscs
A-nnonymous Dec 15, 2025
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
Prev Previous commit
fix miscs
  • Loading branch information
@A-nnonymous
A-nnonymous committed Dec 15, 2025
commit a1b3234db36c83146d99407780203681c08fc4b1
27 changes: 13 additions & 14 deletions paddle/phi/kernels/impl/matmul_grad_kernel_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -283,7 +283,6 @@ void MatmulGradKernel(const Context& dev_ctx,
is_x_been_broadcasted = x_ndim < y_ndim;
is_y_been_broadcasted = !is_x_been_broadcasted;
} else {
#pragma unroll
for (int i = 0; i < ndim; i++) {
if (x_dims[i] != y_dims[i]) {
is_x_been_broadcasted = x_dims[i] < y_dims[i];
Expand Down Expand Up @@ -463,15 +462,15 @@ void MatmulGradKernel(const Context& dev_ctx,
phi::TransposeLast2Dim<T>(dev_ctx, out_grad);
DenseTensor y_conj_processed =
phi::TransposeLast2Dim<T>(dev_ctx, y_conj);
size_t BN = 1;
int64_t BN = 1;
for (int i = 0; i < ndim - 1; i++) {
BN *= y_dims[i];
}
std::vector<std::int64_t> out_grad_2d_dim{BN, dout_dims[ndim - 2]};
std::vector<std::int64_t> y_conj_2d_dim{BN, y_dims[y_ndim - 2]};

out_grad_processed->Resize(common::make_ddim(out_grad_2d_dim));
y_conj_processed->Resize(common::make_ddim(y_conj_2d_dim));
out_grad_processed.Resize(common::make_ddim(out_grad_2d_dim));
y_conj_processed.Resize(common::make_ddim(y_conj_2d_dim));
// 2D x 2D -> 2D
MatMulFunction<Context, T>(dev_ctx,
out_grad_processed,
Expand All @@ -487,9 +486,9 @@ void MatmulGradKernel(const Context& dev_ctx,
for (int i = 0; i < ndim - 2; i++) {
x_grad_dim[i] = 1;
}
x_grad_dims[ndim - 2] = dx_help.dims()[0];
x_grad_dims[ndim - 1] = dx_help.dims()[1];
dx_help->Resize(common::make_ddim(x_grad_dim));
x_grad_dim[ndim - 2] = dx_help.dims()[0];
x_grad_dim[ndim - 1] = dx_help.dims()[1];
dx_help.Resize(common::make_ddim(x_grad_dim));

} else {
MatMulFunction<Context, T>(dev_ctx,
Expand All @@ -507,7 +506,7 @@ void MatmulGradKernel(const Context& dev_ctx,
// Once y been broadcasted, we introduce a new aggregate dim
// original: [B, M, K] x [B, M, N] -> [B, K, N] -(reduceB)-> [K, N]
// new: [BM, K]' x [BM, N] -> [K, N]
size_t BM = 1;
int64_t BM = 1;
for (int i = 0; i < ndim - 1; i++) {
BM *= x_dims[i];
}
Expand All @@ -516,8 +515,8 @@ void MatmulGradKernel(const Context& dev_ctx,

DenseTensor out_grad_processed = out_grad;
DenseTensor x_conj_processed = x_conj;
out_grad_processed->Resize(common::make_ddim(out_grad_2d_dim));
x_conj_processed->Resize(common::make_ddim(x_conj_2d_dim));
out_grad_processed.Resize(common::make_ddim(out_grad_2d_dim));
x_conj_processed.Resize(common::make_ddim(x_conj_2d_dim));

MatMulFunction<Context, T>(dev_ctx,
x_conj_processed,
Expand All @@ -528,13 +527,13 @@ void MatmulGradKernel(const Context& dev_ctx,
true,
false);
// make legacy reduce logic happy
std::vector<std::int64_t> x_grad_dim(ndim);
std::vector<std::int64_t> y_grad_dim(ndim);
for (int i = 0; i < ndim - 2; i++) {
y_grad_dim[i] = 1;
}
y_grad_dims[ndim - 2] = dy_help.dims()[0];
y_grad_dims[ndim - 1] = dy_help.dims()[1];
dy_help->Resize(common::make_ddim(y_grad_dim));
y_grad_dim[ndim - 2] = dy_help.dims()[0];
y_grad_dim[ndim - 1] = dy_help.dims()[1];
dy_help.Resize(common::make_ddim(y_grad_dim));

} else {
MatMulFunction<Context, T>(dev_ctx,
Expand Down
Loading