SummarizationTask

class flash.text.seq2seq.summarization.model.SummarizationTask(backbone='sshleifer/distilbart-xsum-1-1', tokenizer_kwargs=None, max_source_length=128, max_target_length=128, padding='max_length', loss_fn=None, optimizer='Adam', lr_scheduler=None, metrics=None, learning_rate=None, num_beams=4, use_stemmer=True, enable_ort=False)

The SummarizationTask is a Task for Seq2Seq text summarization. For more details, see Summarization.

You can change the backbone to any summarization model from HuggingFace/transformers using the backbone argument.

Parameters

• backbone (str) – backbone model to use for the task.

• max_source_length (int) – The maximum length to pad / truncate input sequences to.

• max_target_length (int) – The maximum length to pad / truncate target sequences to.

• padding (Union[str, bool]) – The type of padding to apply. One of: “longest” or True, “max_length”, “do_not_pad” or False.

• loss_fn (Optional[TypeVar(LOSS_FN_TYPE, Callable, Mapping, Sequence, None)]) – Loss function for training.

• optimizer (TypeVar(OPTIMIZER_TYPE, str, Callable, Tuple[str, Dict[str, Any]], None)) – Optimizer to use for training.

• lr_scheduler (Optional[TypeVar(LR_SCHEDULER_TYPE, str, Callable, Tuple[str, Dict[str, Any]], Tuple[str, Dict[str, Any], Dict[str, Any]], None)]) – The LR scheduler to use during training.

• metrics (Optional[TypeVar(METRICS_TYPE, Metric, Mapping, Sequence, None)]) – Metrics to compute for training and evaluation. Defauls to calculating the ROUGE metric. Changing this argument currently has no effect.

• learning_rate (Optional[float]) – Learning rate to use for training, defaults to 3e-4

• num_beams (Optional[int]) – Number of beams to use in validation when generating predictions. Defaults to 4

• use_stemmer (bool) – Whether Porter stemmer should be used to strip word suffixes to improve matching.

• enable_ort (bool) – Enable Torch ONNX Runtime Optimization: https://onnxruntime.ai/docs/#onnx-runtime-for-training

classmethod available_finetuning_strategies(cls)

Returns a list containing the keys of the available Finetuning Strategies.

Return type

List[str]

classmethod available_lr_schedulers(cls)

Returns a list containing the keys of the available LR schedulers.

Return type

List[str]

classmethod available_optimizers(cls)

Returns a list containing the keys of the available Optimizers.

Return type

List[str]

classmethod available_outputs(cls)

Returns the list of available outputs (that can be used during prediction or serving) for this Task.

Examples

..testsetup:

>>> from flash import Task

>>> print(Task.available_outputs())
['preds', 'raw']

Return type

List[str]