CudaMultiHeadAttentionOp.ixx File Reference

Implementation of the CUDA-based Multi-Head Attention operation for transformer models. More...

#include <vector>
#include <memory>
#include <string>
#include <stdexcept>
#include "Kernels/CudaOps.h"
import Compute.CudaDevice;
import Dnn.Modules.Attention;
import Dnn.Tensor;
import Compute.DeviceContext;
import Dnn.TensorTraits;
import Compute.OperationRegistry;
import Compute.UnaryOperation;
import Compute.OperationBase;
import Compute.CudaMemoryResource;
import Compute.DeviceType;
import Compute.OperationType;
import Compute.OperationAttributes;
import Dnn.ComponentConfig;
import Compute.MemoryResource;

Classes
struct	Mila::Dnn::Compute::Detail::cuda_mha_impl< float >
struct	Mila::Dnn::Compute::Detail::cuda_mha_impl< half >
class	Mila::Dnn::Compute::CudaMultiHeadAttentionOp< TInput, TOutput >
	CUDA implementation of the Multi-Head Attention operation for transformer models. More...
class	Mila::Dnn::Compute::CudaMultiHeadAttentionOpRegistrar
	Class responsible for registering the CudaMultiHeadAttentionOp operation. More...

Namespaces
namespace	Mila
namespace	Mila::Dnn
namespace	Mila::Dnn::Compute
namespace	Mila::Dnn::Compute::Detail
	Namespace for CUDA layer normalization implementation details.

Detailed Description

Implementation of the CUDA-based Multi-Head Attention operation for transformer models.