Skip to content

[Metax] Fix Bug #2472

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
metax666 merged 3 commits into from
Mar 31, 2026
Merged

[Metax] Fix Bug #2472

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
126c6d3
Remove debug print statements in flash_attn_grad_kernel
metax666 Mar 25, 2026
bab9d00
[Metax] Fix Bug (#224)
duqimeng Mar 31, 2026
76206f8
Merge branch 'develop' into develop
metax666 Mar 31, 2026
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
2 changes: 1 addition & 1 deletion backends/metax_gpu/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -290,7 +290,7 @@ file(
${PADDLE_SOURCE_DIR}/paddle/phi/kernels/unsqueeze_kernel.cc
${PADDLE_SOURCE_DIR}/paddle/phi/kernels/squeeze_grad_kernel.cc
${PADDLE_SOURCE_DIR}/paddle/phi/kernels/squeeze_kernel.cc
${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/sign_kernel.cu.cc
${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/sign_kernel.cu
${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/split_kernel.cu
${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/scatter_nd_add_kernel.cu
${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/scatter_nd_add_grad_kernel.cu
Expand Down
8 changes: 4 additions & 4 deletions backends/metax_gpu/kernels/cuda_kernels/sign_kernel_register.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,7 @@ PD_CUSTOM_KERNEL_REGISTER(sign,
int64_t,
float,
double,
phi::dtype::float16,
phi::dtype::bfloat16,
phi::dtype::complex<float>,
phi::dtype::complex<double>) {}
phi::float16,
phi::bfloat16,
phi::complex64,
phi::complex128) {}
16 changes: 8 additions & 8 deletions backends/metax_gpu/kernels/custom_kernel/flash_attn_grad_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,7 +48,7 @@ void FlashAttnUnpaddedGradKernel_(
DenseTensor* dk,
DenseTensor* dv) {
#ifdef PADDLE_WITH_FLASHATTN
printf("************ FlashAttnUnpaddedGradKernel_ ****************\n");
// printf("************ FlashAttnUnpaddedGradKernel_ ****************\n");
ctx.template Alloc<T>(dq);

DenseTensor dk_tmp;
Expand Down Expand Up @@ -214,7 +214,7 @@ void FlashAttnGradKernel(const Context& ctx,
ctx.template Alloc<T>(dq);
DenseTensor dk_tmp;
if (dk) {
printf("dk input pointer is not nullptr!\n");
// printf("dk input pointer is not nullptr!\n");
if (!is_mha) {
dk_tmp.Resize({batch_size, seqlen_k, num_heads, head_size_og});
ctx.template Alloc<T>(&dk_tmp);
Expand All @@ -224,7 +224,7 @@ void FlashAttnGradKernel(const Context& ctx,
}

} else {
printf("dk input pointer is nullptr!\n");
// printf("dk input pointer is nullptr!\n");
dk_tmp = EmptyLike<T, Context>(ctx, k);
}

Expand All @@ -242,9 +242,9 @@ void FlashAttnGradKernel(const Context& ctx,
}

if (dk_tmp.meta().is_contiguous()) {
printf("dk_tmp alloc memory is contiguous!\n");
// printf("dk_tmp alloc memory is contiguous!\n");
} else {
printf("dk_tmp alloc memory is not contiguous!\n");
// printf("dk_tmp alloc memory is not contiguous!\n");
}

FlashAttnParamsBwd params = FlashAttnParamsBwd(ctx,
Expand Down Expand Up @@ -278,7 +278,7 @@ void FlashAttnGradKernel(const Context& ctx,
VLOG(10) << "[FlashAttn Backward] attn_mask.shape=["
<< (attn_mask.get_ptr())->dims() << "]";
}
// printf("params.dq dims[2]:%d, params.dk dims[2]:%d, params.dv
// // printf("params.dq dims[2]:%d, params.dk dims[2]:%d, params.dv
// dims[2]:%d\n", params.dq->head_num, params.dk->head_num,
// params.dv->head_num);
print_tensor_info(params.dq);
Expand Down Expand Up @@ -326,7 +326,7 @@ void FlashAttnGradKernel(const Context& ctx,
phi::SumKernel<T, Context>(ctx, dk_tmp, {3}, dk->type(), false, dk);
} else {
// kvReduceBatchedForGQA<T, Context>(ctx, dk_tmp, dk);
printf("[%s: %d] - Need to complete!\n", __func__, __LINE__);
// printf("[%s: %d] - Need to complete!\n", __func__, __LINE__);
}
}

Expand All @@ -340,7 +340,7 @@ void FlashAttnGradKernel(const Context& ctx,
phi::SumKernel<T, Context>(ctx, dv_tmp, {3}, dv->type(), false, dv);
} else {
// kvReduceBatchedForGQA<T, Context>(ctx, dv_tmp, dv);
printf("[%s: %d] - Need to complete!\n", __func__, __LINE__);
// printf("[%s: %d] - Need to complete!\n", __func__, __LINE__);
}
}
}
Expand Down
Loading
Loading