Skip to content

Steps — Load

Load step: puts a dataset into the results.

Load a dataset into the pipeline results.

Writes to results:

  • results['dataset']: MuranoDataset or LabeledDataset
  • results['prompts']: PromptBatch derived from the dataset texts

Args:

  • dataset: Contrastive or labeled dataset to make available to
  • downstream steps.
def __init__(self, dataset: MuranoDataset | LabeledDataset):
def expected_write_types(self, results = None, available_types = None):

Return {"dataset": <constructor dataset's type>, "prompts": PromptBatch}.