Method __init__

cdslib/core/nn/modules/conv.py:300–578

Multiple Conv1D layers with normalization and nonlinearity. Args: num_layers (int): total number of layers of the network dim_input (int): input dimension dim_features (int or list of int): an integ

(
        self,
        num_layers: int,
        dim_input: int,
        dim_output: int,
        dim_features: T.Union[T.List[int], int],
        kernel_sizes: T.Union[T.List[int], int],
        strides: T.Union[T.List[int], int] = 1,
        paddings: T.Union[T.List[int], int] = 0,
        dilations: T.Union[T.List[int], int] = 1,
        groups: T.Union[T.List[int], int] = 1,
        padding_modes: T.Union[T.List[str], str] = "zeros",
        nonlinearity: str = "leaky_relu",
        add_norm_layer: bool = True,
        norm_fun=nn.LayerNorm,
        dropout_prob: float = 0.0,
        output_add_nonlinearity: bool = False,
    )

    """

    def __init__(
        self,
        num_layers: int,
        dim_input: int,
        dim_output: int,
        dim_features: T.Union[T.List[int], int],
        kernel_sizes: T.Union[T.List[int], int],
        strides: T.Union[T.List[int], int] = 1,
        paddings: T.Union[T.List[int], int] = 0,
        dilations: T.Union[T.List[int], int] = 1,
        groups: T.Union[T.List[int], int] = 1,
        padding_modes: T.Union[T.List[str], str] = "zeros",
        nonlinearity: str = "leaky_relu",
        add_norm_layer: bool = True,
        norm_fun=nn.LayerNorm,
        dropout_prob: float = 0.0,
        output_add_nonlinearity: bool = False,
    ):
        """
        Multiple Conv1D layers with normalization and nonlinearity.

        Args:
            num_layers (int):
                total number of layers of the network
            dim_input (int):
                input dimension
            dim_features (int or list of int):
                an integer if all layers share the same dim_feature,
                or list of length num_layer-1 (one for each layer except the last layer).
            kernel_sizes (int or list of int):
                an integer if all layers share the same kernel_size,
                or list of length num_layer (one for each layer).
            strides (int or list of int):
                an integer if all layers share the same stride,
                or list of length num_layer (one for each layer).
            paddings (int or list of int):
                an integer if all layers share the same padding,
                or list of length num_layer (one for each layer).
            dilations (int or list of int):
                an integer if all layers share the same dilation,
                or list of length num_layer (one for each layer).
            groups (int or list of int):
                an integer if all layers share the same group,
                or list of length num_layer (one for each layer).
            padding_modes (str or list of str):
                a str if all layers share the same padding_mode,
                or list of length num_layer (one for each layer)
            nonlinearity (str):
                nonlinearity used in-between layers:
                ``'leaky_relu'``, ``'relu'``, ``'tanh'``, ``'sigmoid'``, ``'silu'`` (torch >= 1.7.0)
            add_norm_layer (bool):
                whether to add normalization layers in between linear layers.
            norm_fun:
                function of the normalization
                (should be a function that takes only the dim_feature,
                pass lambda function if want to change default parameters)
                ex: :code:`lambda x: torch.nn.LayerNorm(x, eps=1e-5, elementwise_affine=False)`
            dropout_prob (float):