Skip to content

GitLab

Switch branch/tag
Find file Normal viewHistoryPermalink
reasonbench_ppl_b4a005.py 3.83 KB
Newer Older
Skyfall-xzz's avatar
[Feature] Add ReasonBench(Internal) dataset (#577)  
Skyfall-xzz committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
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
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136 
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.reasonbench import ReasonBenchDataset

reasonbench_eval_cfg = dict(
    evaluator=dict(type=AccEvaluator),
    pred_role="BOT",
)

reader_cfgs, infer_cfgs = [], []
for i in range(2, 5):
    choices = ["A", "B", "C", "D"][:i]

    reader_cfgs.append(dict(
    input_columns=["prompt_ppl"] + choices + ["choices"],
    output_column="label")
    )

    infer_cfgs.append(dict(
    prompt_template=dict(
        type=PromptTemplate,
        template={
            str(id):
            dict(
                round=[
                    dict(role="HUMAN", prompt="{prompt_ppl}Answer:"),
                    dict(role="BOT", prompt=f"{choice}")
                ], )
            for id, choice in enumerate(choices)
        }),
    retriever=dict(type=ZeroRetriever),
    inferencer=dict(type=PPLInferencer)
    ))

CausalReasoningDataset = [
    dict(
        abbr="reasonbench-causal",
        type=ReasonBenchDataset,
        path="data/reasonbench/causal.jsonl",
        reader_cfg=reader_cfgs[0],
        infer_cfg=infer_cfgs[0],
        eval_cfg=reasonbench_eval_cfg),
]

CommonsenseReasoningDataset = [
    dict(
        abbr="reasonbench-commonsense",
        type=ReasonBenchDataset,
        path="data/reasonbench/commonsense.jsonl",
        reader_cfg=reader_cfgs[1],
        infer_cfg=infer_cfgs[1],
        eval_cfg=reasonbench_eval_cfg),
]

AbductiveReasoningDataset = [
    dict(
        abbr="reasonbench-abductive",
        type=ReasonBenchDataset,
        path="data/reasonbench/abductive.jsonl",
        reader_cfg=reader_cfgs[0],
        infer_cfg=infer_cfgs[0],
        eval_cfg=reasonbench_eval_cfg),
]

DeductiveReasoningDataset = [
    dict(
        abbr="reasonbench-deductive",
        type=ReasonBenchDataset,
        path="data/reasonbench/deductive.jsonl",
        reader_cfg=reader_cfgs[1],
        infer_cfg=infer_cfgs[1],
        eval_cfg=reasonbench_eval_cfg),
]

InductiveReasoningDataset = [
    dict(
        abbr="reasonbench-inductive",
        type=ReasonBenchDataset,
        path="data/reasonbench/inductive.jsonl",
        reader_cfg=reader_cfgs[0],
        infer_cfg=infer_cfgs[0],
        eval_cfg=reasonbench_eval_cfg),
]

SymbolicReasoningDataset = [
    dict(
        abbr="reasonbench-symbolic",
        type=ReasonBenchDataset,
        path="data/reasonbench/symbolic.jsonl",
        reader_cfg=reader_cfgs[2],
        infer_cfg=infer_cfgs[2],
        eval_cfg=reasonbench_eval_cfg),
]

CLEVA_CommonsenseReasoningDataset = [
    dict(
        abbr="reasonbench-cleva_commonsense",
        type=ReasonBenchDataset,
        path="data/reasonbench/cleva_commonsense.jsonl",
        reader_cfg=reader_cfgs[1],
        infer_cfg=infer_cfgs[1],
        eval_cfg=reasonbench_eval_cfg),
]

CLEVA_DeductiveReasoningDataset = [
    dict(
        abbr="reasonbench-cleva_deductive",
        type=ReasonBenchDataset,
        path="data/reasonbench/cleva_deductive.jsonl",
        reader_cfg=reader_cfgs[1],
        infer_cfg=infer_cfgs[1],
        eval_cfg=reasonbench_eval_cfg),
]

CLEVA_InductiveReasoningDataset = [
    dict(
        abbr="reasonbench-cleva_inductive",
        type=ReasonBenchDataset,
        path="data/reasonbench/cleva_inductive.jsonl",
        reader_cfg=reader_cfgs[0],
        infer_cfg=infer_cfgs[0],
        eval_cfg=reasonbench_eval_cfg),
]

reasonbench_datasets = \
    CLEVA_CommonsenseReasoningDataset + \
    CLEVA_DeductiveReasoningDataset + \
    CLEVA_InductiveReasoningDataset + \
    CausalReasoningDataset + \
    CommonsenseReasoningDataset + \
    AbductiveReasoningDataset + \
    DeductiveReasoningDataset + \
    InductiveReasoningDataset + \
    SymbolicReasoningDataset