diff --git a/deeptab/architectures/enode.py b/deeptab/architectures/enode.py index fbb8849a..baea66ce 100644 --- a/deeptab/architectures/enode.py +++ b/deeptab/architectures/enode.py @@ -16,10 +16,9 @@ class ENODE(BaseModel): Parameters ---------- - cat_feature_info : dict - Dictionary containing information about categorical features, including their names and dimensions. - num_feature_info : dict - Dictionary containing information about numerical features, including their names and dimensions. + feature_information : tuple + A tuple containing feature information for numerical, categorical, and + embedding features. num_classes : int, optional The number of output classes or target dimensions for regression, by default 1. config : ENODEConfig, optional @@ -30,10 +29,6 @@ class ENODE(BaseModel): Attributes ---------- - cat_feature_info : dict - Stores categorical feature information. - num_feature_info : dict - Stores numerical feature information. use_embeddings : bool Flag indicating if embeddings should be used for categorical and numerical features. embedding_layer : EmbeddingLayer, optional diff --git a/deeptab/architectures/ft_transformer.py b/deeptab/architectures/ft_transformer.py index 6fa3a0da..34181e7f 100644 --- a/deeptab/architectures/ft_transformer.py +++ b/deeptab/architectures/ft_transformer.py @@ -16,10 +16,9 @@ class FTTransformer(BaseModel): Parameters ---------- - cat_feature_info : dict - Dictionary containing information about categorical features, including their names and dimensions. - num_feature_info : dict - Dictionary containing information about numerical features, including their names and dimensions. + feature_information : tuple + A tuple containing feature information for numerical, categorical, and + embedding features. num_classes : int, optional The number of output classes or target dimensions for regression, by default 1. config : FTTransformerConfig, optional @@ -32,10 +31,6 @@ class FTTransformer(BaseModel): ---------- pooling_method : str The pooling method to aggregate features after transformer encoding. - cat_feature_info : dict - Stores categorical feature information. - num_feature_info : dict - Stores numerical feature information. embedding_layer : EmbeddingLayer Layer for embedding categorical and numerical features. norm_f : nn.Module diff --git a/deeptab/architectures/mambatab.py b/deeptab/architectures/mambatab.py index bed7cae3..da05a133 100644 --- a/deeptab/architectures/mambatab.py +++ b/deeptab/architectures/mambatab.py @@ -16,10 +16,9 @@ class MambaTab(BaseModel): Parameters ---------- - cat_feature_info : dict - Dictionary containing information about categorical features, including their names and dimensions. - num_feature_info : dict - Dictionary containing information about numerical features, including their names and dimensions. + feature_information : tuple + A tuple containing feature information for numerical, categorical, and + embedding features. num_classes : int, optional The number of output classes or target dimensions for regression, by default 1. config : MambaTabConfig, optional @@ -30,10 +29,6 @@ class MambaTab(BaseModel): Attributes ---------- - cat_feature_info : dict - Stores categorical feature information. - num_feature_info : dict - Stores numerical feature information. initial_layer : nn.Linear Linear layer for the initial transformation of concatenated feature embeddings. norm_f : LayerNorm diff --git a/deeptab/architectures/mambattention.py b/deeptab/architectures/mambattention.py index 709b01e6..13e5b0ad 100644 --- a/deeptab/architectures/mambattention.py +++ b/deeptab/architectures/mambattention.py @@ -16,10 +16,9 @@ class MambAttention(BaseModel): Parameters ---------- - cat_feature_info : dict - Dictionary containing information about categorical features, including their names and dimensions. - num_feature_info : dict - Dictionary containing information about numerical features, including their names and dimensions. + feature_information : tuple + A tuple containing feature information for numerical, categorical, and + embedding features. num_classes : int, optional The number of output classes or target dimensions for regression, by default 1. config : MambAttentionConfig, optional diff --git a/deeptab/architectures/mambular.py b/deeptab/architectures/mambular.py index be1bf593..64dbd5e0 100644 --- a/deeptab/architectures/mambular.py +++ b/deeptab/architectures/mambular.py @@ -15,10 +15,9 @@ class Mambular(BaseModel): Parameters ---------- - cat_feature_info : dict - Dictionary containing information about categorical features, including their names and dimensions. - num_feature_info : dict - Dictionary containing information about numerical features, including their names and dimensions. + feature_information : tuple + A tuple containing feature information for numerical, categorical, and + embedding features. num_classes : int, optional The number of output classes or target dimensions for regression, by default 1. config : MambularConfig, optional @@ -53,7 +52,7 @@ class Mambular(BaseModel): def __init__( self, - feature_information: tuple, # Expecting (cat_feature_info, num_feature_info, embedding_feature_info) + feature_information: tuple, # Expecting (num_feature_info, cat_feature_info, embedding_feature_info) num_classes=1, config: MambularConfig = MambularConfig(), # noqa: B008 **kwargs, diff --git a/deeptab/architectures/mlp.py b/deeptab/architectures/mlp.py index 918f9e3f..6107c47e 100644 --- a/deeptab/architectures/mlp.py +++ b/deeptab/architectures/mlp.py @@ -15,10 +15,9 @@ class MLP(BaseModel): Parameters ---------- - cat_feature_info : dict - Dictionary containing information about categorical features, including their names and dimensions. - num_feature_info : dict - Dictionary containing information about numerical features, including their names and dimensions. + feature_information : tuple + A tuple containing feature information for numerical, categorical, and + embedding features. num_classes : int, optional The number of output classes or target dimensions for regression, by default 1. config : MLPConfig, optional @@ -31,10 +30,6 @@ class MLP(BaseModel): ---------- layer_sizes : list of int List specifying the number of units in each layer of the MLP. - cat_feature_info : dict - Stores categorical feature information. - num_feature_info : dict - Stores numerical feature information. layers : nn.ModuleList List containing the layers of the MLP, including linear layers, normalization layers, and activations. skip_connections : bool diff --git a/deeptab/architectures/ndtf.py b/deeptab/architectures/ndtf.py index 476f6ff7..91cd7c12 100644 --- a/deeptab/architectures/ndtf.py +++ b/deeptab/architectures/ndtf.py @@ -15,10 +15,9 @@ class NDTF(BaseModel): Parameters ---------- - cat_feature_info : dict - Dictionary containing information about categorical features, including their names and dimensions. - num_feature_info : dict - Dictionary containing information about numerical features, including their names and dimensions. + feature_information : tuple + A tuple containing feature information for numerical, categorical, and + embedding features. num_classes : int, optional The number of output classes or target dimensions for regression, by default 1. config : NDTFConfig, optional @@ -30,10 +29,6 @@ class NDTF(BaseModel): Attributes ---------- - cat_feature_info : dict - Stores categorical feature information. - num_feature_info : dict - Stores numerical feature information. penalty_factor : float Scaling factor for the penalty applied during training, specified in the self.hparams. input_dimensions : list of int diff --git a/deeptab/architectures/node.py b/deeptab/architectures/node.py index 4b515500..e8df5fd2 100644 --- a/deeptab/architectures/node.py +++ b/deeptab/architectures/node.py @@ -16,10 +16,9 @@ class NODE(BaseModel): Parameters ---------- - cat_feature_info : dict - Dictionary containing information about categorical features, including their names and dimensions. - num_feature_info : dict - Dictionary containing information about numerical features, including their names and dimensions. + feature_information : tuple + A tuple containing feature information for numerical, categorical, and + embedding features. num_classes : int, optional The number of output classes or target dimensions for regression, by default 1. config : NODEConfig, optional @@ -30,10 +29,6 @@ class NODE(BaseModel): Attributes ---------- - cat_feature_info : dict - Stores categorical feature information. - num_feature_info : dict - Stores numerical feature information. use_embeddings : bool Flag indicating if embeddings should be used for categorical and numerical features. embedding_layer : EmbeddingLayer, optional diff --git a/deeptab/architectures/resnet.py b/deeptab/architectures/resnet.py index a75c45a7..7ebd1aec 100644 --- a/deeptab/architectures/resnet.py +++ b/deeptab/architectures/resnet.py @@ -15,10 +15,9 @@ class ResNet(BaseModel): Parameters ---------- - cat_feature_info : dict - Dictionary containing information about categorical features, including their names and dimensions. - num_feature_info : dict - Dictionary containing information about numerical features, including their names and dimensions. + feature_information : tuple + A tuple containing feature information for numerical, categorical, and + embedding features. num_classes : int, optional The number of output classes or target dimensions for regression, by default 1. config : ResNetConfig, optional @@ -31,10 +30,6 @@ class ResNet(BaseModel): ---------- layer_sizes : list of int List specifying the number of units in each layer of the ResNet. - cat_feature_info : dict - Stores categorical feature information. - num_feature_info : dict - Stores numerical feature information. activation : callable Activation function used in the residual blocks. use_embeddings : bool diff --git a/deeptab/architectures/saint.py b/deeptab/architectures/saint.py index adb919a6..0ca2ae38 100644 --- a/deeptab/architectures/saint.py +++ b/deeptab/architectures/saint.py @@ -15,10 +15,9 @@ class SAINT(BaseModel): Parameters ---------- - cat_feature_info : dict - Dictionary containing information about categorical features, including their names and dimensions. - num_feature_info : dict - Dictionary containing information about numerical features, including their names and dimensions. + feature_information : tuple + A tuple containing feature information for numerical, categorical, and + embedding features. num_classes : int, optional The number of output classes or target dimensions for regression, by default 1. config : SAINTConfig, optional @@ -31,10 +30,6 @@ class SAINT(BaseModel): ---------- pooling_method : str The pooling method to aggregate features after transformer encoding. - cat_feature_info : dict - Stores categorical feature information. - num_feature_info : dict - Stores numerical feature information. embedding_layer : EmbeddingLayer Layer for embedding categorical and numerical features. norm_f : nn.Module diff --git a/deeptab/architectures/tabtransformer.py b/deeptab/architectures/tabtransformer.py index fff4b299..fde8ed7a 100644 --- a/deeptab/architectures/tabtransformer.py +++ b/deeptab/architectures/tabtransformer.py @@ -17,10 +17,9 @@ class TabTransformer(BaseModel): Parameters ---------- - cat_feature_info : dict - Dictionary containing information about categorical features. - num_feature_info : dict - Dictionary containing information about numerical features. + feature_information : tuple + A tuple containing feature information for numerical, categorical, and + embedding features. num_classes : int, optional Number of output classes (default is 1). config : TabTransformerConfig, optional @@ -40,10 +39,6 @@ class TabTransformer(BaseModel): Factor by which the learning rate will be reduced. pooling_method : str Method to pool the features. - cat_feature_info : dict - Dictionary containing information about categorical features. - num_feature_info : dict - Dictionary containing information about numerical features. embedding_activation : callable Activation function for embeddings. encoder: callable diff --git a/deeptab/configs/core.py b/deeptab/configs/core.py index dcce8255..9470f701 100644 --- a/deeptab/configs/core.py +++ b/deeptab/configs/core.py @@ -100,9 +100,9 @@ class BaseModelConfig(BaseEstimator): This class contains only architectural / structural configuration. Training-related parameters (``lr``, ``weight_decay``, ``max_epochs``, …) - belong in :class:`~deeptab.configs.trainer_config.TrainerConfig`. + belong in :class:`~deeptab.configs.TrainerConfig`. Preprocessing parameters belong in - :class:`~deeptab.configs.preprocessing_config.PreprocessingConfig`. + :class:`~deeptab.configs.PreprocessingConfig`. Parameters ---------- diff --git a/deeptab/configs/experimental/modernnca_config.py b/deeptab/configs/experimental/modernnca_config.py index 35f347dd..5ca2102c 100644 --- a/deeptab/configs/experimental/modernnca_config.py +++ b/deeptab/configs/experimental/modernnca_config.py @@ -13,7 +13,7 @@ class ModernNCAConfig(BaseModelConfig): Parameters ---------- embedding_type : str, default='plr' - Type of feature embedding to use (e.g., 'plr', 'ple'). + Type of feature embedding to use (e.g., 'plr', 'linear'). plr_lite : bool, default=True Whether to use the lightweight PLR embedding variant. n_frequencies : int, default=75 diff --git a/deeptab/configs/experimental/trompt_config.py b/deeptab/configs/experimental/trompt_config.py index 428ff453..07f99118 100644 --- a/deeptab/configs/experimental/trompt_config.py +++ b/deeptab/configs/experimental/trompt_config.py @@ -21,7 +21,7 @@ class TromptConfig(BaseModelConfig): n_cells : int, default=4 Number of cells in each cycle. P : int, default=128 - Number of steps in the Trompt model. + Number of prompts in the Trompt model. """ # Trompt-specific architecture diff --git a/deeptab/configs/models/mambatab_config.py b/deeptab/configs/models/mambatab_config.py index 2e6b1ccc..562a1aca 100644 --- a/deeptab/configs/models/mambatab_config.py +++ b/deeptab/configs/models/mambatab_config.py @@ -27,19 +27,19 @@ class MambaTabConfig(BaseModelConfig): dropout : float, default=0.05 Dropout rate for regularization. dt_rank : str, default='auto' - Rank of the decision tree used in the model. + Rank of the dynamic time components used in the model. d_state : int, default=128 Dimensionality of the state in recurrent layers. dt_scale : float, default=1.0 - Scaling factor for the decision tree. + Scaling factor for the dynamic time components. dt_init : str, default='random' - Initialization method for the decision tree. + Initialization method for the dynamic time components. dt_max : float, default=0.1 - Maximum value for decision tree initialization. + Maximum value for dynamic time component initialization. dt_min : float, default=0.0001 - Minimum value for decision tree initialization. + Minimum value for dynamic time component initialization. dt_init_floor : float, default=0.0001 - Floor value for decision tree initialization. + Floor value for dynamic time component initialization. axis : int, default=1 Axis along which operations are applied, if applicable. head_layer_sizes : list, default=[] diff --git a/deeptab/configs/models/mambattention_config.py b/deeptab/configs/models/mambattention_config.py index c816b44e..4b3b9eed 100644 --- a/deeptab/configs/models/mambattention_config.py +++ b/deeptab/configs/models/mambattention_config.py @@ -37,19 +37,19 @@ class MambAttentionConfig(BaseModelConfig): attn_dropout : float, default=0.2 Dropout rate for the attention mechanism. dt_rank : str, default='auto' - Rank of the decision tree. + Rank of the dynamic time components. d_state : int, default=128 Dimensionality of the state in recurrent layers. dt_scale : float, default=1.0 - Scaling factor for the decision tree. + Scaling factor for the dynamic time components. dt_init : str, default='random' - Initialization method for the decision tree. + Initialization method for the dynamic time components. dt_max : float, default=0.1 - Maximum value for decision tree initialization. + Maximum value for dynamic time component initialization. dt_min : float, default=0.0001 - Minimum value for decision tree initialization. + Minimum value for dynamic time component initialization. dt_init_floor : float, default=0.0001 - Floor value for decision tree initialization. + Floor value for dynamic time component initialization. norm : str, default='LayerNorm' Type of normalization used in the model. AD_weight_decay : bool, default=True diff --git a/deeptab/configs/models/mambular_config.py b/deeptab/configs/models/mambular_config.py index 1a978c9f..21fb6d23 100644 --- a/deeptab/configs/models/mambular_config.py +++ b/deeptab/configs/models/mambular_config.py @@ -29,19 +29,19 @@ class MambularConfig(BaseModelConfig): dropout : float, default=0.0 Dropout rate for regularization. dt_rank : str, default='auto' - Rank of the decision tree used in the model. + Rank of the dynamic time components used in the model. d_state : int, default=128 Dimensionality of the state in recurrent layers. dt_scale : float, default=1.0 - Scaling factor for decision tree parameters. + Scaling factor for dynamic time component parameters. dt_init : str, default='random' - Initialization method for decision tree parameters. + Initialization method for dynamic time component parameters. dt_max : float, default=0.1 - Maximum value for decision tree initialization. + Maximum value for dynamic time component initialization. dt_min : float, default=0.0001 - Minimum value for decision tree initialization. + Minimum value for dynamic time component initialization. dt_init_floor : float, default=0.0001 - Floor value for decision tree initialization. + Floor value for dynamic time component initialization. norm : str, default='RMSNorm' Type of normalization used ('RMSNorm', etc.). conv_bias : bool, default=False diff --git a/deeptab/configs/models/mlp_config.py b/deeptab/configs/models/mlp_config.py index 08576686..3365dbad 100644 --- a/deeptab/configs/models/mlp_config.py +++ b/deeptab/configs/models/mlp_config.py @@ -11,9 +11,9 @@ class MLPConfig(BaseModelConfig): """Architecture-only configuration for MLP models (DeepTab 2.0 API). Contains only structural hyperparameters. Training parameters (``lr``, - ``max_epochs``, …) go in :class:`~deeptab.configs.trainer_config.TrainerConfig` + ``max_epochs``, …) go in :class:`~deeptab.configs.TrainerConfig` and preprocessing parameters go in - :class:`~deeptab.configs.preprocessing_config.PreprocessingConfig`. + :class:`~deeptab.configs.PreprocessingConfig`. Parameters ---------- diff --git a/deeptab/configs/models/tabr_config.py b/deeptab/configs/models/tabr_config.py index 7c3cc0b8..bf84d1fc 100644 --- a/deeptab/configs/models/tabr_config.py +++ b/deeptab/configs/models/tabr_config.py @@ -11,12 +11,12 @@ class TabRConfig(BaseModelConfig): """Architecture-only configuration for TabR models (DeepTab 2.0 API). Training fields (``lr``, ``weight_decay``, ``lr_factor``) are configured - via :class:`~deeptab.configs.trainer_config.TrainerConfig`. + via :class:`~deeptab.configs.TrainerConfig`. Parameters ---------- embedding_type : str, default='plr' - Type of feature embedding to use (e.g., 'plr', 'ple'). + Type of feature embedding to use (e.g., 'plr', 'linear'). plr_lite : bool, default=True Whether to use the lightweight PLR embedding variant. n_frequencies : int, default=75 diff --git a/deeptab/nn/blocks/resnet.py b/deeptab/nn/blocks/resnet.py index b215d424..9ac282fa 100644 --- a/deeptab/nn/blocks/resnet.py +++ b/deeptab/nn/blocks/resnet.py @@ -13,8 +13,8 @@ def __init__(self, input_dim, output_dim, activation, norm=False, dropout=0.0): Output dimension of the block. activation : Callable Activation function. - norm_layer : Callable, optional - Normalization layer function, by default None. + norm : bool, optional + Whether to apply layer normalization, by default False. dropout : float, optional Dropout rate, by default 0.0. """