enerzyme.models.layers.scalar_embedding.ScalarResidualMLPEmbedding

class enerzyme.models.layers.scalar_embedding.ScalarResidualMLPEmbedding(dim_embedding: int, embed_field: str, num_residual: int, activation_fn: Literal['shifted_softplus', 'swish'] | None = 'swish', activation_params: Dict[Literal['dim_feature', 'initial_alpha', 'initial_beta', 'learnable'], int | float | bool] = {}, initial_weight1: Tensor | ndarray | Literal['semi_orthogonal_glorot', 'orthogonal', 'zero', 'xavier_uniform'] = 'orthogonal', initial_weight2: Tensor | ndarray | Literal['semi_orthogonal_glorot', 'orthogonal', 'zero', 'xavier_uniform'] = 'orthogonal', initial_weight_out: Tensor | ndarray | Literal['semi_orthogonal_glorot', 'orthogonal', 'zero', 'xavier_uniform'] = 'orthogonal', initial_bias_residual: Tensor | ndarray | Literal['zero'] = 'zero', initial_bias_out: Tensor | ndarray | Literal['zero'] = 'zero', dropout_rate: float = 0, use_bias_residual: bool = True, use_bias_out: bool = True, use_residual: bool = True)

Bases: ScalarEmbedding

__init__(dim_embedding: int, embed_field: str, num_residual: int, activation_fn: Literal['shifted_softplus', 'swish'] | None = 'swish', activation_params: Dict[Literal['dim_feature', 'initial_alpha', 'initial_beta', 'learnable'], int | float | bool] = {}, initial_weight1: Tensor | ndarray | Literal['semi_orthogonal_glorot', 'orthogonal', 'zero', 'xavier_uniform'] = 'orthogonal', initial_weight2: Tensor | ndarray | Literal['semi_orthogonal_glorot', 'orthogonal', 'zero', 'xavier_uniform'] = 'orthogonal', initial_weight_out: Tensor | ndarray | Literal['semi_orthogonal_glorot', 'orthogonal', 'zero', 'xavier_uniform'] = 'orthogonal', initial_bias_residual: Tensor | ndarray | Literal['zero'] = 'zero', initial_bias_out: Tensor | ndarray | Literal['zero'] = 'zero', dropout_rate: float = 0, use_bias_residual: bool = True, use_bias_out: bool = True, use_residual: bool = True) → None