matgl.data package

This package implements data manipulation tools.

matgl.data.transformer module

Module implementing various data transformers for PyTorch.

class matgl.data.transformer.LogTransformer

Bases: Transformer

Performs a natural log of the data.

inverse_transform(data)

Invert the log (exp).

• Parameters: data – Input data
• Returns: exp(data)

transform(data)

Take the log of the data.

• Parameters: data – Input data
• Returns: Scaled data

class matgl.data.transformer.Normalizer(mean: float, std: float)

Bases: Transformer

Performs a scaling of the data by centering to the mean and dividing by the standard deviation.

• Parameters:
  • mean – Mean of the data
  • std – Standard deviation of the data.

classmethod from_data(data)

Create Normalizer from data.

• Parameters: data – Input data.
• Returns: Normalizer

inverse_transform(data)

Invert the scaling.

• Parameters: data – Scaled data
• Returns: Unscaled data

transform(data)

z-score the data by subtracting the mean and dividing by the standard deviation.

• Parameters: data – Input data
• Returns: Scaled data

class matgl.data.transformer.Transformer

Bases: object

Abstract base class defining a data transformer.

abstract inverse_transform(data: Tensor)

Inverse transformation to be performed on data.

• Parameters: data – Input data
• Returns: Inverse-transformed data.

abstract transform(data: Tensor)

Transformation to be performed on data.

• Parameters: data – Input data
• Returns: Transformed data.