Options for the BatchNorm module. More...

#include <batchnorm.h>

Public Member Functions
	BatchNormOptions (int64_t features)

	TORCH_ARG (int64_t, features)
	The number of features of the input tensor. More...

	TORCH_ARG (bool, affine)
	Whether to learn a scale and bias that are applied in an affine transformation on the input. More...

	TORCH_ARG (bool, stateful)
	Whether to store and update batch statistics (mean and variance) in the module. More...

	TORCH_ARG (double, eps)
	The epsilon value added for numerical stability. More...

	TORCH_ARG (double, momentum)=0.1
	A momentum multiplier for the mean and variance. More...

Detailed Description

Options for the BatchNorm module.

Definition at line 13 of file batchnorm.h.

Member Function Documentation

torch::nn::BatchNormOptions::TORCH_ARG	(	int64_t	,
		features
	)

The number of features of the input tensor.

Changing this parameter after construction has no effect.

torch::nn::BatchNormOptions::TORCH_ARG	(	bool	,
		affine
	)

Whether to learn a scale and bias that are applied in an affine transformation on the input.

Changing this parameter after construction has no effect.

torch::nn::BatchNormOptions::TORCH_ARG	(	bool	,
		stateful
	)

Whether to store and update batch statistics (mean and variance) in the module.

If false, you should call pure_forward and supply those batch statistics yourself. Changing this parameter after construction has no effect.

torch::nn::BatchNormOptions::TORCH_ARG	(	double	,
		eps
	)

The epsilon value added for numerical stability.

Changing this parameter after construction is effective.

torch::nn::BatchNormOptions::TORCH_ARG	(	double	,
		momentum
	)

pure virtual

A momentum multiplier for the mean and variance.

Changing this parameter after construction is effective.

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