cudnn_convolution.h

Go to the documentation of this file.

#ifndef INFINI_OPS_BASE_CUDNN_CONVOLUTION_H_
#define INFINI_OPS_BASE_CUDNN_CONVOLUTION_H_
 
#include <vector>
 
#include "operator.h"
 
namespace infini::ops {
 
class CudnnConvolution : public Operator<CudnnConvolution> {
 public:
  CudnnConvolution(const Tensor input, const Tensor weight,
                   const std::vector<int64_t> padding,
                   const std::vector<int64_t> stride,
                   const std::vector<int64_t> dilation, const int64_t groups,
                   const bool benchmark, const bool deterministic,
                   const bool allow_tf32, Tensor out)
      : input_shape_{input.shape()},
        input_strides_{input.strides()},
        input_type_{input.dtype()},
        weight_shape_{weight.shape()},
        weight_strides_{weight.strides()},
        weight_type_{weight.dtype()},
        out_shape_{out.shape()},
        out_strides_{out.strides()},
        out_type_{out.dtype()},
        padding_{padding},
        stride_{stride},
        dilation_{dilation},
        groups_{groups},
        benchmark_{benchmark},
        deterministic_{deterministic},
        allow_tf32_{allow_tf32},
        device_index_{out.device().index()} {}
 
  virtual void operator()(const Tensor input, const Tensor weight,
                          const std::vector<int64_t> padding,
                          const std::vector<int64_t> stride,
                          const std::vector<int64_t> dilation,
                          const int64_t groups, const bool benchmark,
                          const bool deterministic, const bool allow_tf32,
                          Tensor out) const = 0;
 
 protected:
  Tensor::Shape input_shape_;
 
  Tensor::Strides input_strides_;
 
  DataType input_type_;
 
  Tensor::Shape weight_shape_;
 
  Tensor::Strides weight_strides_;
 
  DataType weight_type_;
 
  Tensor::Shape out_shape_;
 
  Tensor::Strides out_strides_;
 
  DataType out_type_;
 
  std::vector<int64_t> padding_{};
 
  std::vector<int64_t> stride_{};
 
  std::vector<int64_t> dilation_{};
 
  int64_t groups_{};
 
  bool benchmark_{};
 
  bool deterministic_{};
 
  bool allow_tf32_{};
 
  int device_index_{0};
};
 
}  // namespace infini::ops
 
#endif