Trainers

class TabularTrainer

TabularTrainer.__init__

def __init__(model: TabularModel, dp_budget: DpBudget | None = None) -> None

Trainer for a TabularModel.

Arguments:

• model - A TabularModel instance.
• dp_budget - The (eps, delta)-budget for differentially private (DP) training. If None (the default), the training will not be differentially private. Available only for single table datasets.

TabularTrainer.train

def train(dataset: TabularDataset,
          n_epochs: int | None = None,
          n_steps: int | None = None,
          batch_size: int = 0,
          lr: float = 0.0,
          memory: int = 0,
          valid: Validation | None = None,
          hooks: Sequence[TrainHook] = (),
          accumulate_grad: int = 1,
          dp_step: DpStep | None = None) -> None

Train the tabular model with the input dataset.

Arguments:

• dataset - The training data, as a TabularDataset object.
• n_epochs - The number of training epochs. One and only one of n_epochs and n_steps must be provided.
• n_steps - The number of training steps. One and only one of n_epochs and n_steps must be provided.
• batch_size - The size of a batch of data during training. When it is not specified the user must provide the argument memory.
• lr - The learning rate. If it is 0 the optimal value for the learning rate is automatically determined.
• memory - The available memory in MB that is used to automatically compute the optimal value of the batch size.
• valid - A Validation object. If None, no validation is performed.
• hooks - A sequence of custom TrainHook objects.
• accumulate_grad - The number of gradient accumulation steps. If equal to 1, the weights are updated at each step.
• dp_step - Data for differentially private step. Must be provided if and only if the trainer has a DP-budget.

TabularTrainer.save

def save(path: Path | str) -> None

Save the trainer to a checkpoint at the given path.

TabularTrainer.load

@classmethod
def load(cls: type["TabularTrainer"], path: Path | str) -> "TabularTrainer"

Load the trainer from the checkpoint at the given path.

class TextTrainer

TextTrainer.__init__

def __init__(model: TextModel) -> None

Trainer for a TextModel.

Arguments:

• model - A TextModel instance.

TextTrainer.train

def train(dataset: TextDataset,
          n_epochs: int | None = None,
          n_steps: int | None = None,
          batch_size: int = 0,
          lr: float = 0.0,
          memory: int = 0,
          valid: Validation | None = None,
          hooks: Sequence[TrainHook] = (),
          accumulate_grad: int = 1) -> None

Train the text model with the input dataset.

Arguments:

• dataset - The training data, as a TextDataset object.
• n_epochs - The number of training epochs. One and only one of n_epochs and n_steps must be provided.
• n_steps - The number of training steps. One and only one of n_epochs and n_steps must be provided.
• batch_size - The size of a batch of data during training. When it is not specified the user must provide the argument memory.
• lr - The learning rate. If it is 0 the optimal value for the learning rate is automatically determined.
• memory - The available memory in MB that is used to automatically compute the optimal value of the batch size.
• valid - A Validation object. If None, no validation is performed.
• hooks - A sequence of custom TrainHook objects.
• accumulate_grad - The number of gradient accumulation steps. If equal to 1, the weights are updated at each step.

TextTrainer.save

def save(path: Path | str) -> None

Save the trainer to a checkpoint at the given path.

TextTrainer.load

@classmethod
def load(cls: type["TextTrainer"], path: Path | str) -> "TextTrainer"

Load the trainer from the checkpoint at the given path.