File size: 6,024 Bytes
1faccd4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
137
138
139
140
141
142
143
144
145
146
147
148
149
150
# Copyright 2024 Bytedance Ltd. and/or its affiliates
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Preprocess the historical AIME dataset to parquet format.
"""

import argparse
import json
import os

import datasets

from verl.utils.hdfs_io import copy, makedirs


DATA_SOURCE = "gneubig/aime-1983-2024"
INSTRUCTION_FOLLOWING = "Let's think step by step and output the final answer within \\boxed{}."
KEY_CANDIDATES = {
    "year": ["Year", "year"],
    "problem_number": ["Problem Number", "Problem_Number", "problem_number"],
    "question": ["Problem", "Question", "problem", "question"],
    "answer": ["Answer", "answer", "solution"],
}


def find_key(column_names, key_name):
    for candidate in KEY_CANDIDATES[key_name]:
        if candidate in column_names:
            return candidate
    raise KeyError(f"Could not find {key_name} column in dataset columns: {column_names}")


def load_raw_dataset(local_dataset_path=None):
    if local_dataset_path is None:
        dataset = datasets.load_dataset(DATA_SOURCE)
    elif os.path.isdir(local_dataset_path):
        dataset = datasets.load_dataset(local_dataset_path)
    elif local_dataset_path.endswith(".csv"):
        dataset = datasets.load_dataset("csv", data_files=local_dataset_path)
    elif local_dataset_path.endswith(".json") or local_dataset_path.endswith(".jsonl"):
        dataset = datasets.load_dataset("json", data_files=local_dataset_path)
    else:
        dataset = datasets.load_dataset(local_dataset_path)

    if isinstance(dataset, datasets.DatasetDict):
        if "train" in dataset:
            return dataset["train"]
        return next(iter(dataset.values()))

    return dataset


def build_dataset(raw_dataset, split_name, year_min=None, year_max=None, data_source="aime_history"):
    year_key = find_key(raw_dataset.column_names, "year")
    problem_number_key = find_key(raw_dataset.column_names, "problem_number")
    question_key = find_key(raw_dataset.column_names, "question")
    answer_key = find_key(raw_dataset.column_names, "answer")

    def year_in_range(example):
        year = int(example[year_key])
        if year_min is not None and year < year_min:
            return False
        if year_max is not None and year > year_max:
            return False
        return True

    filtered_dataset = raw_dataset.filter(year_in_range)

    def process_fn(example, idx):
        year = int(example[year_key])
        problem_number = int(example[problem_number_key])
        question = str(example[question_key]).strip()
        ground_truth = str(example[answer_key]).strip()
        prompt = f"{question} {INSTRUCTION_FOLLOWING}"
        return {
            "data_source": data_source,
            "prompt": [{"role": "user", "content": prompt}],
            "ability": "math",
            "reward_model": {"style": "rule", "ground_truth": ground_truth},
            "extra_info": {
                "split": split_name,
                "index": idx,
                "year": year,
                "problem_number": problem_number,
                "question": question,
            },
        }

    return filtered_dataset.map(process_fn, with_indices=True, remove_columns=filtered_dataset.column_names)


if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument("--local_dir", default=None)
    parser.add_argument("--hdfs_dir", default=None)
    parser.add_argument("--local_dataset_path", default=None)
    parser.add_argument(
        "--local_save_dir",
        default="~/data/aime",
        help="The save directory for the preprocessed dataset.",
    )
    args = parser.parse_args()

    local_save_dir = args.local_dir
    if local_save_dir is not None:
        print("Warning: Argument 'local_dir' is deprecated. Please use 'local_save_dir' instead.")
    else:
        local_save_dir = args.local_save_dir

    local_dir = os.path.expanduser(local_save_dir)
    os.makedirs(local_dir, exist_ok=True)

    print(f"Loading the {DATA_SOURCE} dataset from huggingface...", flush=True)
    raw_dataset = load_raw_dataset(args.local_dataset_path)

    dataset_all = build_dataset(raw_dataset, split_name="train", year_min=1983, year_max=2024)
    dataset_until_2022 = build_dataset(raw_dataset, split_name="train", year_min=1983, year_max=2022)
    dataset_until_2023 = build_dataset(raw_dataset, split_name="train", year_min=1983, year_max=2023)
    dataset_2023 = build_dataset(raw_dataset, split_name="test", year_min=2023, year_max=2023, data_source="aime23")
    dataset_2024 = build_dataset(raw_dataset, split_name="test", year_min=2024, year_max=2024, data_source="aime24")

    outputs = [
        ("aime-history-1983-2024.parquet", "aime-history-1983-2024.example.json", dataset_all),
        ("aime-history-1983-2022.parquet", "aime-history-1983-2022.example.json", dataset_until_2022),
        ("aime-history-1983-2023.parquet", "aime-history-1983-2023.example.json", dataset_until_2023),
        ("aime-history-2023.parquet", "aime-history-2023.example.json", dataset_2023),
        ("aime-history-2024.parquet", "aime-history-2024.example.json", dataset_2024),
    ]

    for parquet_name, example_name, dataset in outputs:
        dataset.to_parquet(os.path.join(local_dir, parquet_name))
        if len(dataset) > 0:
            with open(os.path.join(local_dir, example_name), "w") as f:
                json.dump(dataset[0], f, indent=2)

    if args.hdfs_dir is not None:
        makedirs(args.hdfs_dir)
        copy(src=local_dir, dst=args.hdfs_dir)