Skip to content

[Big tensor] fix cuda error in margin_cross_entropy #74155

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
merged 1 commit into from
Jul 22, 2025
+7 −5
Merged

[Big tensor] fix cuda error in margin_cross_entropy #74155

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

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 paddle/phi/kernels/gpu/margin_cross_entropy_grad_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,7 +80,7 @@ void MarginCrossEntropyGradKernel(const Context& dev_ctx,
logits_grad->ShareDataWith(softmax);
}

int blocks = NumBlocks(N * D);
int64_t blocks = NumBlocks(N * D);
int threads = kNumCUDAThreads;
const auto& label_type = label.dtype();

Expand Down
10 changes: 6 additions & 4 deletions paddle/phi/kernels/gpu/margin_cross_entropy_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,15 +67,17 @@ __global__ void ScaleLogitKernel(T* logits,
const float scale,
const int64_t N,
const int64_t D) {
CUDA_KERNEL_LOOP(i, N * D) { logits[i] *= static_cast<T>(scale); }
CUDA_KERNEL_LOOP_TYPE(i, N * D, int64_t) {
logits[i] *= static_cast<T>(scale);
}
}

template <typename T>
__global__ void LogitsMinusMaxKernel(T* logits,
const T* logits_max_per_row,
const int64_t N,
const int64_t D) {
CUDA_KERNEL_LOOP(i, N * D) {
CUDA_KERNEL_LOOP_TYPE(i, N * D, int64_t) {
auto row = i / D;
logits[i] -= logits_max_per_row[row];
}
Expand All @@ -86,7 +88,7 @@ __global__ void LogitsMinusLogSumKernel(T* logits,
const T* logits_sum_per_row,
const int64_t N,
const int64_t D) {
CUDA_KERNEL_LOOP(i, N * D) {
CUDA_KERNEL_LOOP_TYPE(i, N * D, int64_t) {
auto row = i / D;
logits[i] -= phi::kps::details::Log(logits_sum_per_row[row]);
}
Expand All @@ -102,7 +104,7 @@ __global__ void HardLabelSoftmaxWithCrossEntropyKernel(
const int64_t D,
const int* class_interval_ptr) {
int start_index = class_interval_ptr[rank];
CUDA_KERNEL_LOOP(i, N * D) {
CUDA_KERNEL_LOOP_TYPE(i, N * D, int64_t) {
auto row = i / D;
auto col = i % D;
if ((col + start_index) == labels[row]) {
Expand Down