nemo_rl.data.datasets.response_datasets.tulu3#

Module Contents#

Classes#

Tulu3SftMixtureDataset

Simple wrapper around the Tulu3 SFT mixture dataset with train split.

API#

class nemo_rl.data.datasets.response_datasets.tulu3.Tulu3SftMixtureDataset(
split_validation_size: float = 0.05,
seed: int = 42,
max_samples: int | None = None,
**kwargs,
)#

Bases: nemo_rl.data.datasets.raw_dataset.RawDataset

Simple wrapper around the Tulu3 SFT mixture dataset with train split.

Parameters:

  • split_validation_size – Size of the validation data, default is 0.05

  • seed – Seed for train/validation split when split_validation_size > 0, default is 42

  • max_samples – Optional maximum number of samples to use from the dataset

Initialization

format_data(data: dict[str, Any]) → dict[str, Any]#