tlc.integration.hugging_face.trainer¶
Module Contents¶
Classes¶
Class |
Description |
|---|---|
API¶
- class Trainer(
- model: PreTrainedModel | Module | None = None,
- args: TrainingArguments | None = None,
- data_collator: Callable | None = None,
- train_dataset: Table | None = None,
- eval_dataset: Table | dict[str, Table] | None = None,
- processing_class: PreTrainedTokenizerBase | BaseImageProcessor | FeatureExtractionMixin | ProcessorMixin | None = None,
- model_init: Callable[[], PreTrainedModel] | None = None,
- compute_loss_func: Callable | None = None,
- compute_metrics: Callable[[EvalPrediction], dict] | None = None,
- callbacks: list[TrainerCallback] | None = None,
- optimizers: tuple[Optimizer | None, LambdaLR | None] = (None, None),
- preprocess_logits_for_metrics: Callable[[Tensor, Tensor], Tensor] | None = None,
- *,
- run_name: str | None = None,
- run_description: str | None = None,
- exclude_zero_weights_metrics_collection: bool = False,
- exclude_zero_weights_train: bool = True,
- weighted: bool = True,
- shuffle: bool = True,
- repeat_by_weight: bool = False,
- metrics_collectors: tlc.client.torch.metrics.metrics_collectors.metrics_collector_base.MetricsCollectorType | None = None,
- metrics_collection_epochs: list[int] | None = None,
Bases:
transformers.Trainer