hellaswag_ppl_a6e128.py 1.31 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
from opencompass.openicl.icl_prompt_template import PromptTemplate
from opencompass.openicl.icl_retriever import ZeroRetriever
from opencompass.openicl.icl_inferencer import PPLInferencer
from opencompass.openicl.icl_evaluator import AccEvaluator
from opencompass.datasets import hellaswagDataset_V3

hellaswag_reader_cfg = dict(
    input_columns=['query', 'A', 'B', 'C', 'D'],
    output_column='gold')

hellaswag_infer_cfg = dict(
    prompt_template=dict(
        type=PromptTemplate,
        template={
15
16
            '0': dict(
                round=[dict(role='HUMAN', prompt='{query} {A}')]
17
            ),
18
19
            '1': dict(
                round=[dict(role='HUMAN', prompt='{query} {B}')]
20
            ),
21
22
            '2': dict(
                round=[dict(role='HUMAN', prompt='{query} {C}')]
23
            ),
24
25
            '3': dict(
                round=[dict(role='HUMAN', prompt='{query} {D}')]
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
            ),
        }),
    retriever=dict(type=ZeroRetriever),
    inferencer=dict(type=PPLInferencer))

hellaswag_eval_cfg = dict(evaluator=dict(type=AccEvaluator))

hellaswag_datasets = [
    dict(
        abbr='hellaswag',
        type=hellaswagDataset_V3,
        path='./data/hellaswag/hellaswag.jsonl',
        reader_cfg=hellaswag_reader_cfg,
        infer_cfg=hellaswag_infer_cfg,
        eval_cfg=hellaswag_eval_cfg)
]