faenet.force_decoder

Module Contents

Classes

ForceDecoder	Predicts a force vector per atom from final atomic representations.
LambdaLayer

class faenet.force_decoder.ForceDecoder(type, input_channels, model_configs, act)

Bases: torch.nn.Module

Predicts a force vector per atom from final atomic representations.

Parameters:

• type (str) – Type of force decoder to use

• input_channels (int) – Number of input channels

• model_configs (dict) – Dictionary of config parameters for the decoder’s model

• act (callable) – Activation function (NOT a module)

Raises:

ValueError – Unknown type of decoder

Returns:

Predicted force vector per atom

Return type:

(torch.Tensor)

forward(h)

reset_parameters()

class faenet.force_decoder.LambdaLayer(func)

Bases: torch.nn.Module

forward(x)