Skip to content
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

Jump to bottom

fix error due to () not used on operator priority. #14699

Merged
merged 1 commit into from
Feb 16, 2023
Merged

fix error due to () not used on operator priority. #14699

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
9 changes: 5 additions & 4 deletions onnxruntime/contrib_ops/cuda/bert/relative_attn_bias.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -146,7 +146,8 @@ Status GatedRelativePositionBias<T>::ComputeInternal(OpKernelContext* context) c
const auto BNS = batch_size * num_heads_ * seq_len;
const size_t elements_in_query = (size_t)BNS * (size_t)head_size;
const size_t elements_after_gemm = (size_t)BNS *(size_t)D;
size_t workspace_size = sizeof(T) * (elements_in_query + (seq_len < D) ? elements_after_gemm : (size_t)0);
bool reuse_output = (seq_len >= D);
size_t workspace_size = sizeof(T) * (elements_in_query + (reuse_output ? (size_t)0 : elements_after_gemm));
auto workspace = GetScratchBuffer<void>(workspace_size, context->GetComputeStream());

// format 1: BxSx(NH * total_matrix) => matrix_to_transpose * (BxNxSxH)
Expand All @@ -161,9 +162,9 @@ Status GatedRelativePositionBias<T>::ComputeInternal(OpKernelContext* context) c
false, head_size, reinterpret_cast<CudaT*>(static_cast<CudaT*>(nullptr)), total_maxtrix);

// reuse output if possible
CudaT* gemm_output = (seq_len < D) ? (reinterpret_cast<CudaT*>(workspace.get()) + elements_in_query)
: reinterpret_cast<CudaT*>(output->template MutableData<T>());
int ld_gemm_output = max(seq_len, D);
CudaT* gemm_output = reuse_output ? reinterpret_cast<CudaT*>(output->template MutableData<T>())
: (reinterpret_cast<CudaT*>(workspace.get()) + elements_in_query);
int ld_gemm_output = reuse_output ? seq_len : D;

const CudaT one = ToCudaType<T>::FromFloat(1.0f);
const CudaT zero = ToCudaType<T>::FromFloat(0.0f);
Expand Down