Skip to content

[BIG tensor] fix overflow in cudnn_softmax by replacing the api #74329

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
wanghuancoder merged 2 commits into from
Aug 1, 2025
Merged

[BIG tensor] fix overflow in cudnn_softmax by replacing the api #74329

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4d219a0
[BIG tensor] fix overflow in cudnn_softmax by replacing the api
hxzd5568 Jul 31, 2025
b4df84e
[BIG tensor] minor fix
hxzd5568 Jul 31, 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
35 changes: 9 additions & 26 deletions paddle/phi/kernels/gpu/cross_entropy_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -713,14 +713,16 @@ template <typename T>
static void SoftmaxWithCrossEntropySoftLabel(const GPUContext& dev_ctx,
const int rank,
const int axis,
const T* logits_data,
const DenseTensor& logits,
const T* labels_data,
T* softmax_data,
DenseTensor* softmax,
T* loss_data,
int N,
int dim,
int D) {
constexpr int kMaxBlockDim = 512;
auto* logits_data = logits.data<T>();
auto* softmax_data = softmax->data<T>();
int64_t block_dim = dim >= kMaxBlockDim
? kMaxBlockDim
: (1 << static_cast<int>(std::log2(dim)));
Expand Down Expand Up @@ -762,13 +764,7 @@ static void SoftmaxWithCrossEntropySoftLabel(const GPUContext& dev_ctx,
GPUDNNDataLayout layout = GPUDNNDataLayout::kNCHW;
#ifdef PADDLE_WITH_HIP
miopenTensorDescriptor_t descp = desc.descriptor<T>(layout, tensor_dims);
#else
cudnnTensorDescriptor_t descp = desc.descriptor<T>(layout, tensor_dims);
#endif

auto handle = dev_ctx.cudnn_handle();

#ifdef PADDLE_WITH_HIP
auto mode = axis == rank - 1 ? MIOPEN_SOFTMAX_MODE_INSTANCE
: MIOPEN_SOFTMAX_MODE_CHANNEL;
PADDLE_ENFORCE_GPU_SUCCESS(phi::dynload::miopenSoftmaxForward_V2(
Expand All @@ -782,18 +778,8 @@ static void SoftmaxWithCrossEntropySoftLabel(const GPUContext& dev_ctx,
MIOPEN_SOFTMAX_LOG,
mode));
#else
auto mode = axis == rank - 1 ? CUDNN_SOFTMAX_MODE_INSTANCE
: CUDNN_SOFTMAX_MODE_CHANNEL;
PADDLE_ENFORCE_GPU_SUCCESS(phi::dynload::cudnnSoftmaxForward(
handle,
CUDNN_SOFTMAX_LOG,
mode,
phi::backends::gpu::CudnnDataType<T>::kOne(),
descp,
logits_data,
phi::backends::gpu::CudnnDataType<T>::kZero(),
descp,
softmax_data));
SoftmaxForwardCUDAKernelDriver<T, true>(dev_ctx, logits, axis, softmax);
softmax_data = softmax->data<T>();
#endif

const int kDimLog2 = static_cast<int>(Log2Ceil(dim));
Expand Down Expand Up @@ -1170,7 +1156,7 @@ static void SoftmaxWithCrossEntropyHardLabel(const GPUContext& dev_ctx,
VLOG(7) << "rank=" << rank << ", axis = " << axis << ", N = " << N
<< ", dim = " << dim << ", D = " << D;
auto* logits_data = logits.data<T>();
auto* softmax_data = dev_ctx.template Alloc<T>(softmax);
auto* softmax_data = softmax->data<T>();
auto stream = dev_ctx.stream();
constexpr int max_dim = 320;
if (D == 1) {
Expand Down Expand Up @@ -1216,8 +1202,6 @@ static void SoftmaxWithCrossEntropyHardLabel(const GPUContext& dev_ctx,
MIOPEN_SOFTMAX_LOG,
mode));
#else
auto mode = axis == rank - 1 ? CUDNN_SOFTMAX_MODE_INSTANCE
: CUDNN_SOFTMAX_MODE_CHANNEL;
SoftmaxForwardCUDAKernelDriver<T, true>(dev_ctx, logits, axis, softmax);
softmax_data = softmax->data<T>();
#endif
Expand Down Expand Up @@ -1352,14 +1336,13 @@ void CrossEntropyWithSoftmaxCUDAKernel(const GPUContext& dev_ctx,
}

if (soft_label) {
auto* logits_data = logits.data<T>();
auto* labels_data = label.data<T>();
SoftmaxWithCrossEntropySoftLabel<T>(dev_ctx,
rank,
axis_v,
logits_data,
logits,
labels_data,
softmax_data,
softmax,
loss_data,
n,
axis_dim,
Expand Down
Loading