Conv2D Class Reference

Conv2D allows to create a forward convolution primitive. More...

#include <layers_fwd.h>

Public Member Functions
	Conv2D (int batch_size, int patch_length, int n_kernels, int kernel_size, int stride_length, int padding_length, int dilation, dnnl::memory input, std::vector< dnnl::primitive > &net, std::vector< std::unordered_map< int, dnnl::memory >> &net_args, dnnl::engine eng)
	Construct a new Conv 2 D object. More...

Public Attributes
dnnl::memory	arg_src
	Source memory handler.
dnnl::memory	arg_dst
	Destination memory handler.
dnnl::memory	arg_bias
	Bias memory handler.
dnnl::memory	arg_weights
	Weights memory handler.

Detailed Description

Conv2D allows to create a forward convolution primitive.

Constructor & Destructor Documentation

Conv2D()

Conv2D::Conv2D	(	int	batch_size,
		int	patch_length,
		int	n_kernels,
		int	kernel_size,
		int	stride_length,
		int	padding_length,
		int	dilation,
		dnnl::memory	input,
		std::vector< dnnl::primitive > &	net,
		std::vector< std::unordered_map< int, dnnl::memory >> &	net_args,
		dnnl::engine	eng
	)

Construct a new Conv 2 D object.

Parameters

batch_size	Size of the batch
patch_length	Length of the H and W
n_kernels	Number of kernels
kernel_size	Size of the kernel
stride_length	Stride
padding_length	Padding
dilation	Dilation coefficient for the dilated convolution (0 for no dilation as per oneAPI specs)
input	Input memory
net	This is the vector of primitives to which we will append the FC layer primitive
net_args	This is the associated map to which we will add the arguments of the primitive
eng	oneAPI engine that will host the primitive

---

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

• src/layers/include/layers_fwd.h