dragon-ml-toolbox 19.13.0__py3-none-any.whl → 20.0.0__py3-none-any.whl

ml_tools/_core/_ML_models_pytab.py

@@ -1,693 +0,0 @@
-import torch
-from torch import nn
-from torch.utils.data import DataLoader
-from typing import Union, Dict, Any, Literal
-from pathlib import Path
-import json
-import warnings
-
-from ._ML_models import _ArchitectureHandlerMixin
-from ._path_manager import make_fullpath
-from ._keys import PytorchModelArchitectureKeys
-from ._schema import FeatureSchema
-from ._script_info import _script_info
-from ._logger import get_logger
-
-
-_LOGGER = get_logger("Pytorch Tabular")
-
-
-# Imports from pytorch_tabular
-try:
-    from omegaconf import DictConfig
-    from pytorch_tabular.models import (
-        GatedAdditiveTreeEnsembleModel as _GATE, 
-        NODEModel as _NODE,
-        TabNetModel as _TabNet,
-        AutoIntModel as _AutoInt
-    )
-except ImportError:
-    _LOGGER.error(f"GATE and NODE require 'pip install pytorch_tabular omegaconf' dependencies.")
-    raise ImportError()
-else:
-    # Silence pytorch_tabular INFO logs up to error level
-    import logging
-    logging.getLogger("pytorch_tabular").setLevel(logging.ERROR)
-    logging.getLogger("pytorch_tabular.models.node.node_model").setLevel(logging.ERROR)
-
-
-__all__ = [
-    "PyTabGateModel",
-    "PyTabTabNet",
-    "PyTabAutoInt",
-    "PyTabNodeModel"
-]
-
-
-class _BasePytabWrapper(nn.Module, _ArchitectureHandlerMixin):
-    """
-    Internal Base Class: Do not use directly.
-    
-    This is an adapter to make pytorch_tabular models compatible with the
-    dragon-ml-toolbox pipeline.
-    """
-    def __init__(self, schema: FeatureSchema):
-        super().__init__()
-        
-        self.schema = schema
-        self.model_name = "Base" # To be overridden by child
-        self.internal_model: nn.Module = None # type: ignore # To be set by child
-        self.model_hparams: Dict = dict() # To be set by child
-
-        # --- Derive indices from schema ---
-        categorical_map = schema.categorical_index_map
-        
-        if categorical_map:
-            # The order of keys/values is implicitly linked and must be preserved
-            self.categorical_indices = list(categorical_map.keys())
-            self.cardinalities = list(categorical_map.values())
-        else:
-            self.categorical_indices = []
-            self.cardinalities = []
-        
-        # Derive numerical indices by finding what's not categorical
-        all_indices = set(range(len(schema.feature_names)))
-        categorical_indices_set = set(self.categorical_indices)
-        self.numerical_indices = sorted(list(all_indices - categorical_indices_set))
-
-    def _build_pt_config(self, out_targets: int, **kwargs) -> DictConfig:
-        """Helper to create the minimal config dict for a pytorch_tabular model."""
-        task = "regression"
-
-        config_dict = {
-            # --- Data / Schema Params ---
-            'task': task,
-            'continuous_cols': list(self.schema.continuous_feature_names),
-            'categorical_cols': list(self.schema.categorical_feature_names),
-            'continuous_dim': len(self.numerical_indices),
-            'categorical_dim': len(self.categorical_indices),
-            'categorical_cardinality': self.cardinalities,
-            'target': ['dummy_target'], # Required, but not used
-            
-            # --- Model Params ---
-            'output_dim': out_targets,
-            'target_range': None,
-            **kwargs
-        }
-        
-        if 'loss' not in config_dict:
-            config_dict['loss'] = 'MSELoss' # Dummy
-        if 'metrics' not in config_dict:
-            config_dict['metrics'] = []
-            
-        return DictConfig(config_dict)
-    
-    def _build_inferred_config(self, out_targets: int, embedding_dim: int = None) -> DictConfig:
-        """
-        Helper to create the inferred_config required by pytorch_tabular v1.0+.
-        Includes explicit embedding_dims calculation to satisfy BaseModel assertions.
-        """
-        # 1. Calculate embedding_dims list of tuples: [(cardinality, dim), ...]
-        if self.categorical_indices:
-            if embedding_dim is not None:
-                # Use the user-provided fixed dimension for all categorical features
-                embedding_dims = [(card, embedding_dim) for card in self.cardinalities]
-            else:
-                # Default heuristic: min(50, (card + 1) // 2)
-                embedding_dims = [(card, min(50, (card + 1) // 2)) for card in self.cardinalities]
-        else:
-            embedding_dims = []
-
-        # 2. Calculate the total dimension of concatenated embeddings
-        # This fixes the 'Missing key embedded_cat_dim' error
-        embedded_cat_dim = sum([dim for _, dim in embedding_dims])
-
-        return DictConfig({
-            "continuous_dim": len(self.numerical_indices),
-            "categorical_dim": len(self.categorical_indices),
-            "categorical_cardinality": self.cardinalities,
-            "output_dim": out_targets,
-            "embedding_dims": embedding_dims,
-            "embedded_cat_dim": embedded_cat_dim, 
-        })
-
-    def forward(self, x: torch.Tensor) -> torch.Tensor:
-        """
-        Accepts a single tensor and converts it to the dict
-        that pytorch_tabular models expect.
-        """
-        x_cont = x[:, self.numerical_indices].float()
-        x_cat = x[:, self.categorical_indices].long()
-
-        input_dict = {
-            'continuous': x_cont,
-            'categorical': x_cat
-        }
-        
-        model_output_dict = self.internal_model(input_dict)
-        return model_output_dict['logits']
-
-    def get_architecture_config(self) -> Dict[str, Any]:
-        """Returns the full configuration of the model."""
-        schema_dict = {
-            'feature_names': self.schema.feature_names,
-            'continuous_feature_names': self.schema.continuous_feature_names,
-            'categorical_feature_names': self.schema.categorical_feature_names,
-            'categorical_index_map': self.schema.categorical_index_map,
-            'categorical_mappings': self.schema.categorical_mappings
-        }
-        
-        config = {
-            'schema_dict': schema_dict,
-            'out_targets': self.out_targets,
-            **self.model_hparams
-        }
-        return config
-
-    @classmethod
-    def load(cls: type, file_or_dir: Union[str, Path], verbose: bool = True) -> nn.Module:
-        """Loads a model architecture from a JSON file."""
-        user_path = make_fullpath(file_or_dir)
-        
-        if user_path.is_dir():
-            json_filename = PytorchModelArchitectureKeys.SAVENAME + ".json"
-            target_path = make_fullpath(user_path / json_filename, enforce="file")
-        elif user_path.is_file():
-            target_path = user_path
-        else:
-            _LOGGER.error(f"Invalid path: '{file_or_dir}'")
-            raise IOError()
-
-        with open(target_path, 'r') as f:
-            saved_data = json.load(f)
-
-        saved_class_name = saved_data[PytorchModelArchitectureKeys.MODEL]
-        config = saved_data[PytorchModelArchitectureKeys.CONFIG]
-
-        if saved_class_name != cls.__name__:
-            _LOGGER.error(f"Model class mismatch. File specifies '{saved_class_name}', but '{cls.__name__}' was expected.")
-            raise ValueError()
-
-        # --- RECONSTRUCTION LOGIC ---
-        if 'schema_dict' not in config:
-            _LOGGER.error("Invalid architecture file: missing 'schema_dict'.")
-            raise ValueError("Missing 'schema_dict' in config.")
-            
-        schema_data = config.pop('schema_dict')
-        
-        raw_index_map = schema_data['categorical_index_map']
-        if raw_index_map is not None:
-            rehydrated_index_map = {int(k): v for k, v in raw_index_map.items()}
-        else:
-            rehydrated_index_map = None
-
-        schema = FeatureSchema(
-            feature_names=tuple(schema_data['feature_names']),
-            continuous_feature_names=tuple(schema_data['continuous_feature_names']),
-            categorical_feature_names=tuple(schema_data['categorical_feature_names']),
-            categorical_index_map=rehydrated_index_map,
-            categorical_mappings=schema_data['categorical_mappings']
-        )
-        
-        config['schema'] = schema
-        # --- End Reconstruction ---
-
-        model = cls(**config)
-        if verbose:
-            _LOGGER.info(f"Successfully loaded architecture for '{saved_class_name}'")
-        return model
-    
-    def __repr__(self) -> str:
-        internal_model_str = str(self.internal_model)
-        internal_repr = internal_model_str.split('\n')[0]
-        return f"{self.model_name}(internal_model={internal_repr})"
-
-
-class PyTabGateModel(_BasePytabWrapper):
-    """
-    Adapter for the Gated Additive Tree Ensemble (GATE) model.
-    """
-    def __init__(self, *,
-                 schema: FeatureSchema,
-                 out_targets: int,
-                 embedding_dim: int = 32,
-                 gflu_stages: int = 4,
-                 num_trees: int = 20,
-                 tree_depth: int = 4,
-                 dropout: float = 0.1):
-        """
-        Args:
-            schema (FeatureSchema): 
-                The definitive schema object from data_exploration.
-            out_targets (int): 
-                Number of output targets.
-            embedding_dim (int):
-                Dimension of the categorical embeddings. (Recommended: 16 to 64)
-            gflu_stages (int):
-                Number of Gated Feature Learning Units (GFLU) stages. (Recommended: 2 to 6)
-            num_trees (int):
-                Number of trees in the ensemble. (Recommended: 10 to 50)
-            tree_depth (int):
-                Depth of each tree. (Recommended: 4 to 6)
-            dropout (float):
-                Dropout rate for the GFLU.
-        """
-        super().__init__(schema)
-        
-        warnings.filterwarnings("ignore", message="Implicit dimension choice for softmax")
-        warnings.filterwarnings("ignore", message="Ignoring head config")
-        
-        
-        self.model_name = "PyTabGateModel"
-        self.out_targets = out_targets
-        
-        self.model_hparams = {
-            'embedding_dim': embedding_dim,
-            'gflu_stages': gflu_stages,
-            'num_trees': num_trees,
-            'tree_depth': tree_depth,
-            'dropout': dropout
-        }
-
-        # Build Hyperparameter Config with defaults
-        pt_config = self._build_pt_config(
-            out_targets=out_targets,
-            embedding_dim=embedding_dim,
-            
-            # GATE Specific Mappings
-            gflu_stages=gflu_stages,
-            num_trees=num_trees,
-            tree_depth=tree_depth,
-            gflu_dropout=dropout,
-            tree_dropout=dropout,
-            tree_wise_attention=True,
-            tree_wise_attention_dropout=dropout,
-            
-            # GATE Defaults
-            chain_trees=False,
-            binning_activation="sigmoid",
-            feature_mask_function="softmax",
-            share_head_weights=True,
-            
-            # Sparsity
-            gflu_feature_init_sparsity=0.3,
-            tree_feature_init_sparsity=0.3,
-            learnable_sparsity=True,
-            
-            # Head Configuration
-            head="LinearHead",
-            head_config={
-                "layers": "", 
-                "activation": "ReLU", 
-                "dropout": 0.0, 
-                "use_batch_norm": False, 
-                "initialization": "kaiming"
-            },
-            
-            # General Defaults (Required to prevent initialization errors)
-            embedding_dropout=0.0,
-            batch_norm_continuous_input=False,
-            virtual_batch_size=None,
-            learning_rate=1e-3, 
-            target_range=None,
-        )
-        
-        # Build Data Inference Config (Required by PyTabular v1.0+)
-        inferred_config = self._build_inferred_config(
-            out_targets=out_targets, 
-            embedding_dim=embedding_dim
-        )
-
-        # Instantiate the internal pytorch_tabular model
-        self.internal_model = _GATE(
-            config=pt_config, 
-            inferred_config=inferred_config
-        )
-        
-    def __repr__(self) -> str:
-        return (f"{self.model_name}(\n"
-                f"  out_targets={self.out_targets},\n"
-                f"  embedding_dim={self.model_hparams.get('embedding_dim')},\n"
-                f"  gflu_stages={self.model_hparams.get('gflu_stages')},\n"
-                f"  num_trees={self.model_hparams.get('num_trees')},\n"
-                f"  tree_depth={self.model_hparams.get('tree_depth')},\n"
-                f"  dropout={self.model_hparams.get('dropout')}\n"
-                ")")
-
-
-class PyTabTabNet(_BasePytabWrapper):
-    """
-    Adapter for Google's TabNet (Attentive Interpretable Tabular Learning).
-    
-    TabNet uses sequential attention to choose which features to reason 
-    from at each decision step, enabling interpretability.
-    """
-    def __init__(self, *,
-                 schema: FeatureSchema,
-                 out_targets: int,
-                 n_d: int = 8,
-                 n_a: int = 8,
-                 n_steps: int = 3,
-                 gamma: float = 1.3,
-                 n_independent: int = 2,
-                 n_shared: int = 2,
-                 virtual_batch_size: int = 128,
-                 mask_type: Literal['sparsemax', 'entmax', 'softmax'] = 'sparsemax'):
-        """
-        Args:
-            schema (FeatureSchema): The definitive schema object.
-            out_targets (int): Number of output targets.
-            n_d (int): Dimension of the prediction layer (usually 8-64).
-            n_a (int): Dimension of the attention layer (usually equal to n_d).
-            n_steps (int): Number of sequential attention steps (usually 3-10).
-            gamma (float): Relaxation parameter for sparsity (usually 1.0-2.0).
-            n_independent (int): Number of independent GLU layers in each block.
-            n_shared (int): Number of shared GLU layers in each block.
-            virtual_batch_size (int): Batch size for Ghost Batch Normalization.
-            mask_type (str): Masking function.
-                - 'sparsemax' for sparse feature selection.
-                - 'entmax' for moderately sparse selection.
-                - 'softmax' for dense selection (safest).
-        """
-        super().__init__(schema)
-        self.model_name = "PyTabTabNet"
-        self.out_targets = out_targets
-        
-        self.model_hparams = {
-            'n_d': n_d,
-            'n_a': n_a,
-            'n_steps': n_steps,
-            'gamma': gamma,
-            'n_independent': n_independent,
-            'n_shared': n_shared,
-            'virtual_batch_size': virtual_batch_size,
-            'mask_type': mask_type
-        }
-
-        # TabNet does not use standard embeddings, so we don't pass embedding_dim
-        pt_config = self._build_pt_config(
-            out_targets=out_targets,
-            
-            # TabNet Params
-            n_d=n_d,
-            n_a=n_a,
-            n_steps=n_steps,
-            gamma=gamma,
-            n_independent=n_independent,
-            n_shared=n_shared,
-            virtual_batch_size=virtual_batch_size,
-            
-            # TabNet Defaults
-            mask_type=mask_type,
-            
-            # Head Configuration
-            head="LinearHead",
-            head_config={
-                "layers": "", 
-                "activation": "ReLU", 
-                "dropout": 0.0, 
-                "use_batch_norm": False, 
-                "initialization": "kaiming"
-            },
-            
-            # General Defaults
-            batch_norm_continuous_input=False,
-            learning_rate=1e-3
-        )
-        
-        inferred_config = self._build_inferred_config(out_targets=out_targets)
-
-        self.internal_model = _TabNet(
-            config=pt_config,
-            inferred_config=inferred_config
-        )
-
-    def __repr__(self) -> str:
-        return (f"{self.model_name}(\n"
-                f"  out_targets={self.out_targets},\n"
-                f"  n_d={self.model_hparams.get('n_d')},\n"
-                f"  n_a={self.model_hparams.get('n_a')},\n"
-                f"  n_steps={self.model_hparams.get('n_steps')},\n"
-                f"  gamma={self.model_hparams.get('gamma')},\n"
-                f"  virtual_batch_size={self.model_hparams.get('virtual_batch_size')}\n"
-                f"  mask_type='{self.model_hparams.get('mask_type')}'\n"
-                f")")
-
-
-class PyTabAutoInt(_BasePytabWrapper):
-    """
-    Adapter for AutoInt (Automatic Feature Interaction Learning).
-    
-    Uses Multi-Head Self-Attention to automatically learn high-order 
-    feature interactions.
-    """
-    def __init__(self, *,
-                 schema: FeatureSchema,
-                 out_targets: int,
-                 embedding_dim: int = 32,
-                 num_heads: int = 2,
-                 num_attn_blocks: int = 3,
-                 attn_dropout: float = 0.1,
-                 has_residuals: bool = True,
-                 deep_layers: bool = True,
-                 layers: str = "128-64-32"):
-        """
-        Args:
-            schema (FeatureSchema): The definitive schema object.
-            out_targets (int): Number of output targets.
-            embedding_dim (int): Dimension of feature embeddings (attn_embed_dim).
-            num_heads (int): Number of attention heads.
-            num_attn_blocks (int): Number of attention layers.
-            attn_dropout (float): Dropout between attention layers.
-            has_residuals (bool): If True, adds residual connections.
-            deep_layers (bool): If True, adds a standard MLP after attention.
-            layers (str): Hyphen-separated layer sizes for the deep MLP part.
-        """
-        super().__init__(schema)
-        self.model_name = "PyTabAutoInt"
-        self.out_targets = out_targets
-        
-        self.model_hparams = {
-            'embedding_dim': embedding_dim,
-            'num_heads': num_heads,
-            'num_attn_blocks': num_attn_blocks,
-            'attn_dropout': attn_dropout,
-            'has_residuals': has_residuals,
-            'deep_layers': deep_layers,
-            'layers': layers
-        }
-
-        pt_config = self._build_pt_config(
-            out_targets=out_targets,
-            
-            # AutoInt Params
-            attn_embed_dim=embedding_dim,
-            num_heads=num_heads,
-            num_attn_blocks=num_attn_blocks,
-            attn_dropouts=attn_dropout,
-            has_residuals=has_residuals,
-            
-            # Deep MLP part (Optional in AutoInt, but usually good)
-            deep_layers=deep_layers,
-            layers=layers,
-            activation="ReLU",
-            
-            # Head Configuration
-            head="LinearHead",
-            head_config={
-                "layers": "", 
-                "activation": "ReLU", 
-                "dropout": 0.0, 
-                "use_batch_norm": False, 
-                "initialization": "kaiming"
-            },
-            
-            # General Defaults
-            embedding_dropout=0.0,
-            batch_norm_continuous_input=False,
-            learning_rate=1e-3
-        )
-        
-        inferred_config = self._build_inferred_config(
-            out_targets=out_targets, 
-            embedding_dim=embedding_dim
-        )
-
-        self.internal_model = _AutoInt(
-            config=pt_config,
-            inferred_config=inferred_config
-        )
-
-    def __repr__(self) -> str:
-        return (f"{self.model_name}(\n"
-                f"  out_targets={self.out_targets},\n"
-                f"  embedding_dim={self.model_hparams.get('embedding_dim')},\n"
-                f"  num_heads={self.model_hparams.get('num_heads')},\n"
-                f"  num_attn_blocks={self.model_hparams.get('num_attn_blocks')},\n"
-                f"  deep_layers={self.model_hparams.get('deep_layers')}\n"
-                f")")
-
-
-class PyTabNodeModel(_BasePytabWrapper):
-    """
-    Adapter for the Neural Oblivious Decision Ensembles (NODE) model.
-    """
-    def __init__(self, *,
-                 schema: FeatureSchema,
-                 out_targets: int,
-                 embedding_dim: int = 32,
-                 num_trees: int = 1024,
-                 num_layers: int = 2,
-                 tree_depth: int = 6,
-                 dropout: float = 0.1,
-                 backend_function: Literal['softmax', 'entmax15'] = 'softmax'):
-        """
-        Args:
-            schema (FeatureSchema): 
-                The definitive schema object from data_exploration.
-            out_targets (int): 
-                Number of output targets.
-            embedding_dim (int):
-                Dimension of the categorical embeddings. (Recommended: 16 to 64)
-            num_trees (int):
-                Total number of trees in the ensemble. (Recommended: 256 to 2048)
-            num_layers (int):
-                Number of NODE layers (stacked ensembles). (Recommended: 2 to 4)
-            tree_depth (int):
-                Depth of each tree. (Recommended: 4 to 8)
-            dropout (float):
-                Dropout rate.
-            backend_function ('softmax' | 'entmax15'):
-                Function for feature selection. 'entmax15' (sparse) or 'softmax' (dense).
-                Use 'softmax' if dealing with convergence issues.
-        """
-        super().__init__(schema)
-        self.model_name = "PyTabNodeModel"
-        self.out_targets = out_targets
-        
-        warnings.filterwarnings("ignore", message="Ignoring head config because NODE has a specific head")
-
-        self.model_hparams = {
-            'embedding_dim': embedding_dim,
-            'num_trees': num_trees,
-            'num_layers': num_layers,
-            'tree_depth': tree_depth,
-            'dropout': dropout,
-            'backend_function': backend_function
-        }
-
-        # Build Hyperparameter Config with ALL defaults
-        pt_config = self._build_pt_config(
-            out_targets=out_targets,
-            embedding_dim=embedding_dim,
-            
-            # NODE Specific Mappings
-            num_trees=num_trees,
-            depth=tree_depth, # Map tree_depth -> depth
-            num_layers=num_layers,     # num_layers=1 for a single ensemble
-            total_trees=num_trees,
-            dropout_rate=dropout,
-            
-            # NODE Defaults (Manually populated to satisfy backbone requirements)
-            additional_tree_output_dim=0,
-            input_dropout=0.0,
-            choice_function=backend_function,
-            bin_function=backend_function,
-            initialize_response="normal",
-            initialize_selection_logits="uniform",
-            threshold_init_beta=1.0,
-            threshold_init_cutoff=1.0,
-            max_features=None,
-            
-            # General Defaults (Required to prevent initialization errors)
-            embedding_dropout=0.0,
-            batch_norm_continuous_input=False,
-            virtual_batch_size=None,
-            learning_rate=1e-3, 
-            
-            # NODE schema
-            data_aware_init_batch_size=2000, # Required by NodeConfig schema
-            augment_dim=0,
-        )
-
-        # Build Data Inference Config (Required by PyTabular v1.0+)
-        inferred_config = self._build_inferred_config(
-            out_targets=out_targets, 
-            embedding_dim=embedding_dim
-        )
-
-        # Instantiate the internal pytorch_tabular model
-        self.internal_model = _NODE(
-            config=pt_config,
-            inferred_config=inferred_config
-        )
-    
-    def perform_data_aware_initialization(self, train_dataset: Any, batch_size: int = 2000):
-        """
-        CRITICAL: Initializes NODE decision thresholds using a batch of data.
-        
-        Call this ONCE before training starts with a large batch (e.g., 2000 samples).
-        
-        Use the CPU
-        
-        Args:
-            train_dataset: a PyTorch Dataset.
-            batch_size: Number of samples to use for initialization.
-        """
-        # Use a DataLoader to robustly fetch a single batch
-        loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
-        
-        try:
-            batch = next(iter(loader))
-        except StopIteration:
-            _LOGGER.error("Dataset is empty. Cannot perform data-aware initialization.")
-            return
-
-        x_tensor, _ = batch 
-        
-        # Prepare input dict
-        # Prepare input dict matching pytorch_tabular expectations
-        # Ensure we are on the same device as the model (CPU here)
-        device = next(self.parameters()).device
-        x_cont = x_tensor[:, self.numerical_indices].float().to(device)
-        x_cat = x_tensor[:, self.categorical_indices].long().to(device)
-        
-        input_dict = {
-            'continuous': x_cont,
-            'categorical': x_cat
-        }
-        
-        # --- MOCK DATA MODULE ---
-        # datamodule.train_dataloader() -> yields the batch
-        class _MockDataModule:
-            def train_dataloader(self, batch_size=None):
-                # Accepts 'batch_size' argument to satisfy the caller
-                # Returns a list containing just the single pre-processed batch dictionary
-                return [input_dict]
-        
-        mock_dm = _MockDataModule()
-        
-        _LOGGER.info(f"Running NODE Data-Aware Initialization with {batch_size} samples...")
-        try:
-            with torch.no_grad():
-                # Call init on the BACKBONE, not the wrapper
-                self.internal_model.data_aware_initialization(mock_dm)
-            _LOGGER.info("NODE Initialization Complete. Ready to train.")
-        except Exception as e:
-            _LOGGER.error(f"Failed to initialize NODE model: {e}")
-            raise e
-    
-    def __repr__(self) -> str:
-        return (f"{self.model_name}(\n"
-                f"  out_targets={self.out_targets},\n"
-                f"  embedding_dim={self.model_hparams.get('embedding_dim')},\n"
-                f"  num_trees={self.model_hparams.get('num_trees')},\n"
-                f"  num_layers={self.model_hparams.get('num_layers')},\n"
-                f"  tree_depth={self.model_hparams.get('tree_depth')},\n"
-                f"  dropout={self.model_hparams.get('dropout')}\n"
-                f"  backend_function={self.model_hparams.get('backend_function')}\n"
-                f")")
-
-
-def info():
-    _script_info(__all__)