Skip to content

Fix possible big tensor problem in fused_transpose_split_quant #74964

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
Aug 29, 2025
Merged

Fix possible big tensor problem in fused_transpose_split_quant #74964

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
21 changes: 12 additions & 9 deletions paddle/phi/kernels/fusion/gpu/fused_transpose_split_quant_kernel.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,7 +42,8 @@ __device__ void BlockLoad(const InT* input,
const uint32_t local_off_M = threadIdx.y + i * 16;
const uint32_t off_m = blockIdx.x * 128 + local_off_M;
const uint32_t off_k = blockIdx.y * 128 + threadIdx.x * VecSize;
const size_t offset = off_m * K + off_k;
const size_t offset =
static_cast<size_t>(off_m) * static_cast<size_t>(K) + off_k;

float scale;
if constexpr (need_dequant) {
Expand All @@ -53,15 +54,17 @@ __device__ void BlockLoad(const InT* input,

#pragma unroll
for (uint32_t j = 0; j < 4; j += VecSize) {
const size_t idx = offset + j * 32;
using LoadT = VecType<InT, VecSize>;
LoadT data = *reinterpret_cast<const LoadT*>(input + idx);
if (off_k + j * 32 < K) {
const size_t idx = offset + j * 32;
using LoadT = VecType<InT, VecSize>;
LoadT data = *reinterpret_cast<const LoadT*>(input + idx);
#pragma unroll
for (uint32_t k = 0; k < VecSize; k++) {
if constexpr (need_dequant) {
x[i][j + k] = __float2bfloat16(static_cast<float>(data[k]) * scale);
} else {
x[i][j + k] = (*reinterpret_cast<__nv_bfloat16*>(&data[k]));
for (uint32_t k = 0; k < VecSize; k++) {
if constexpr (need_dequant) {
x[i][j + k] = __float2bfloat16(static_cast<float>(data[k]) * scale);
} else {
x[i][j + k] = (*reinterpret_cast<__nv_bfloat16*>(&data[k]));
}
}
}
}
Expand Down