Mila::Dnn::Compute::Detail::cuda_mha_impl< T > Struct Template Reference

Implementation details for CUDA-based Multi-Head Attention operations. More...

Detailed Description

template<typename T>
struct Mila::Dnn::Compute::Detail::cuda_mha_impl< T >

Implementation details for CUDA-based Multi-Head Attention operations.

This namespace contains specialized implementations of Multi-Head Attention operations for different data types (float, half) using CUDA kernels. The implementations are optimized for NVIDIA GPUs for high-performance computation of attention mechanisms in transformer architectures.

---

The documentation for this struct was generated from the following file:

• /home/runner/work/Mila/Mila/Mila/Src/Dnn/Compute/Operations/Cuda/CudaMultiHeadAttentionOp.ixx