std

Functions:

Name	Description
`log_ratio`	Compute the log of the ratio of max scale to the standard deviations.
`masked_negative_log_mean`	Compute the mean of the elements which remain after masking, instead of including the zeroed out elements.
`negative_log_mean`	Compute the negative log mean loss. Useful as a regularization function for noise layer parameters.

log_ratio ¶

log_ratio(std: Tensor, max_scale: Tensor) -> torch.Tensor

Compute the log of the ratio of max scale to the standard deviations.

Parameters:

Name	Type	Description	Default
`std` ¶	`Tensor`	The standard deviation tensor.	required
`max_scale` ¶	`Tensor`	The maximum scale tensor. This is the maximum value that the standard deviation can take. See `stainedglass_core.noise_layer.parameterizations.ScaledStandardDeviationParameterization`.	required

Returns:

Type	Description
`torch.Tensor`	The log of the ratio of the maximum scale to the standard deviation.

Added in version 0.82.0.

masked_negative_log_mean(
    input: Tensor, mask: Tensor
) -> torch.Tensor

Compute the mean of the elements which remain after masking, instead of including the zeroed out elements.

Parameters:

Name	Type	Description	Default
`input` ¶	`Tensor`	The input tensor.	required
`mask` ¶	`Tensor`	The mask tensor.	required

Returns:

Type	Description
`torch.Tensor`	The negative log mean of unmasked elements.

negative_log_mean(input: Tensor) -> torch.Tensor

Compute the negative log mean loss. Useful as a regularization function for noise layer parameters.

Parameters:

Name	Type	Description	Default
`input` ¶	`Tensor`	The input tensor.	required

Returns:

Type	Description
`torch.Tensor`	The negative log mean loss.