Skip to content

TruthfulQA

get_input_data_model()

Returns TruthfulQA (MC1) input data model.

Source code in synalinks/src/datasets/built_in/truthfulqa.py 
40
41
42
43
@synalinks_export("synalinks.datasets.truthfulqa.get_input_data_model")
def get_input_data_model():
    """Returns TruthfulQA (MC1) input data model."""
    return TruthfulQAQuestion

get_output_data_model()

Returns TruthfulQA (MC1) output data model.

Source code in synalinks/src/datasets/built_in/truthfulqa.py 
46
47
48
49
@synalinks_export("synalinks.datasets.truthfulqa.get_output_data_model")
def get_output_data_model():
    """Returns TruthfulQA (MC1) output data model."""
    return TruthfulQAAnswer

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

Streaming dataset for RL-style training.

Returns:

Type Description
HuggingFaceDataset

A streaming, iterable dataset.
Source code in synalinks/src/datasets/built_in/truthfulqa.py 
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
@synalinks_export("synalinks.datasets.truthfulqa.iterable_dataset")
def iterable_dataset(repeat=1, batch_size=1, limit=None, split="validation"):
    """
    Streaming dataset for RL-style training.

    Returns:
        (HuggingFaceDataset): A streaming, iterable dataset.
    """
    return HuggingFaceDataset(
        path="truthful_qa",
        name="multiple_choice",
        split=split,
        streaming=True,
        input_data_model=TruthfulQAQuestion,
        input_template=_INPUT_TEMPLATE,
        output_data_model=TruthfulQAAnswer,
        output_template=_OUTPUT_TEMPLATE,
        batch_size=batch_size,
        limit=limit,
        repeat=repeat,
    )

load_data(validation_split=0.2)

Load TruthfulQA (MC1, multiple_choice config).

HF ships only a single validation split for the MC1 task, so we deterministically split it into train / test.

Parameters:

Name Type Description Default
validation_split float

Fraction held out for evaluation (default 0.2).

 0.2

Returns:

Type Description
tuple

(x_train, y_train), (x_test, y_test).
Source code in synalinks/src/datasets/built_in/truthfulqa.py 
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
@synalinks_export("synalinks.datasets.truthfulqa.load_data")
def load_data(validation_split=0.2):
    """
    Load TruthfulQA (MC1, ``multiple_choice`` config).

    HF ships only a single ``validation`` split for the MC1 task, so we
    deterministically split it into train / test.

    Args:
        validation_split (float): Fraction held out for evaluation
            (default ``0.2``).

    Returns:
        (tuple): ``(x_train, y_train), (x_test, y_test)``.
    """
    x, y = load_split(
        path="truthful_qa",
        name="multiple_choice",
        split="validation",
        input_data_model=TruthfulQAQuestion,
        input_template=_INPUT_TEMPLATE,
        output_data_model=TruthfulQAAnswer,
        output_template=_OUTPUT_TEMPLATE,
    )
    return split_train_test(x, y, validation_split=validation_split)