Skip to content

HellaSwag

get_input_data_model()

Returns HellaSwag input data model.

Source code in synalinks/src/datasets/built_in/hellaswag.py 
31
32
33
34
@synalinks_export("synalinks.datasets.hellaswag.get_input_data_model")
def get_input_data_model():
    """Returns HellaSwag input data model."""
    return HellaSwagQuestion

get_output_data_model()

Returns HellaSwag output data model.

Source code in synalinks/src/datasets/built_in/hellaswag.py 
37
38
39
40
@synalinks_export("synalinks.datasets.hellaswag.get_output_data_model")
def get_output_data_model():
    """Returns HellaSwag output data model."""
    return HellaSwagAnswer

iterable_dataset(repeat=1, batch_size=1, limit=None, split='train')

Streaming dataset for RL-style training.

Returns:

Type Description
HuggingFaceDataset

A streaming, iterable dataset.
Source code in synalinks/src/datasets/built_in/hellaswag.py 
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
@synalinks_export("synalinks.datasets.hellaswag.iterable_dataset")
def iterable_dataset(repeat=1, batch_size=1, limit=None, split="train"):
    """
    Streaming dataset for RL-style training.

    Returns:
        (HuggingFaceDataset): A streaming, iterable dataset.
    """
    return HuggingFaceDataset(
        path="Rowan/hellaswag",
        split=split,
        streaming=True,
        input_data_model=HellaSwagQuestion,
        input_template=_INPUT_TEMPLATE,
        output_data_model=HellaSwagAnswer,
        output_template=_OUTPUT_TEMPLATE,
        batch_size=batch_size,
        limit=limit,
        repeat=repeat,
    )

load_data()

Load HellaSwag.

HF test split has no public labels, so we use train for training and validation for evaluation.

Returns:

Type Description
tuple

(x_train, y_train), (x_test, y_test).
Source code in synalinks/src/datasets/built_in/hellaswag.py 
54
55
56
57
58
59
60
61
62
63
64
65
66
67
@synalinks_export("synalinks.datasets.hellaswag.load_data")
def load_data():
    """
    Load HellaSwag.

    HF test split has no public labels, so we use ``train`` for training
    and ``validation`` for evaluation.

    Returns:
        (tuple): ``(x_train, y_train), (x_test, y_test)``.
    """
    x_train, y_train = _load("train")
    x_test, y_test = _load("validation")
    return (x_train, y_train), (x_test, y_test)