Datasets:

ArXiv:
License:
holylovenia commited on
Commit
4681170
1 Parent(s): 47e66fc

Upload aya_evaluation_suite.py with huggingface_hub

Browse files
Files changed (1) hide show
  1. aya_evaluation_suite.py +147 -0
aya_evaluation_suite.py ADDED
@@ -0,0 +1,147 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from pathlib import Path
2
+ from typing import Dict, List, Tuple
3
+
4
+ import datasets
5
+ import pandas as pd
6
+
7
+ from seacrowd.utils import schemas
8
+ from seacrowd.utils.configs import SEACrowdConfig
9
+ from seacrowd.utils.constants import Licenses, Tasks
10
+
11
+ _CITATION = """
12
+ @misc{singh2024aya,
13
+ title={Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuning},
14
+ author={Shivalika Singh and Freddie Vargus and Daniel Dsouza and Börje F. Karlsson and
15
+ Abinaya Mahendiran and Wei-Yin Ko and Herumb Shandilya and Jay Patel and Deividas
16
+ Mataciunas and Laura OMahony and Mike Zhang and Ramith Hettiarachchi and Joseph
17
+ Wilson and Marina Machado and Luisa Souza Moura and Dominik Krzemiński and Hakimeh
18
+ Fadaei and Irem Ergün and Ifeoma Okoh and Aisha Alaagib and Oshan Mudannayake and
19
+ Zaid Alyafeai and Vu Minh Chien and Sebastian Ruder and Surya Guthikonda and Emad A.
20
+ Alghamdi and Sebastian Gehrmann and Niklas Muennighoff and Max Bartolo and Julia Kreutzer
21
+ and Ahmet Üstün and Marzieh Fadaee and Sara Hooker},
22
+ year={2024},
23
+ eprint={2402.06619},
24
+ archivePrefix={arXiv},
25
+ primaryClass={cs.CL}
26
+ }
27
+ """
28
+
29
+ _DATASETNAME = "aya_evaluation_suite"
30
+
31
+ _DESCRIPTION = """
32
+ Aya Evaluation Suite contains a total of 26,750 open-ended conversation-style
33
+ prompts to evaluate multilingual open-ended generation quality.
34
+ """
35
+
36
+ _HOMEPAGE = "https://huggingface.co/datasets/CohereForAI/aya_evaluation_suite"
37
+
38
+ _LANGUAGES = ["ceb", "tha", "mya", "zsm", "jav", "ind", "vie", "sun", "ace", "bjn", "khm", "lao", "min"]
39
+
40
+ _LICENSE = Licenses.APACHE_2_0.value
41
+
42
+ _LOCAL = False
43
+
44
+ _URLS = {
45
+ _DATASETNAME: "https://huggingface.co/datasets/CohereForAI/aya_evaluation_suite/resolve/main/dolly_machine_translated/test-00000-of-00001.parquet?download=true",
46
+ }
47
+
48
+ _SUPPORTED_TASKS = [Tasks.INSTRUCTION_TUNING]
49
+
50
+ _SOURCE_VERSION = "1.0.0"
51
+
52
+ _SEACROWD_VERSION = "2024.06.20"
53
+
54
+
55
+ class AyaEvaluationSuiteDataset(datasets.GeneratorBasedBuilder):
56
+ """
57
+ Aya Evaluation Suite contains a total of 26,750 open-ended conversation-style
58
+ prompts to evaluate multilingual open-ended generation quality.
59
+ """
60
+
61
+ SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
62
+ SEACROWD_VERSION = datasets.Version(_SEACROWD_VERSION)
63
+
64
+ BUILDER_CONFIGS = [
65
+ SEACrowdConfig(
66
+ name=f"{_DATASETNAME}_{LANG}_source",
67
+ version=datasets.Version(_SOURCE_VERSION),
68
+ description=f"{_DATASETNAME} {LANG} source schema",
69
+ schema="source",
70
+ subset_id=f"{_DATASETNAME}_{LANG}",
71
+ )
72
+ for LANG in _LANGUAGES
73
+ ] + [
74
+ SEACrowdConfig(
75
+ name=f"{_DATASETNAME}_{LANG}_seacrowd_t2t",
76
+ version=datasets.Version(_SEACROWD_VERSION),
77
+ description=f"{_DATASETNAME} {LANG} SEACrowd schema",
78
+ schema="seacrowd_t2t",
79
+ subset_id=f"{_DATASETNAME}_{LANG}",
80
+ )
81
+ for LANG in _LANGUAGES
82
+ ]
83
+
84
+ DEFAULT_CONFIG_NAME = f"{_DATASETNAME}_ind_source"
85
+
86
+ def _info(self) -> datasets.DatasetInfo:
87
+
88
+ if self.config.schema == "source":
89
+ features = datasets.Features(
90
+ {
91
+ "id": datasets.Value("int64"),
92
+ "inputs": datasets.Value("string"),
93
+ "targets": datasets.Value("string"),
94
+ "language": datasets.Value("string"),
95
+ "script": datasets.Value("string"),
96
+ "source_id": datasets.Value("int64"),
97
+ }
98
+ )
99
+
100
+ elif self.config.schema == "seacrowd_t2t":
101
+ features = schemas.text2text_features
102
+
103
+ return datasets.DatasetInfo(
104
+ description=_DESCRIPTION,
105
+ features=features,
106
+ homepage=_HOMEPAGE,
107
+ license=_LICENSE,
108
+ citation=_CITATION,
109
+ )
110
+
111
+ def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
112
+ """Returns SplitGenerators."""
113
+
114
+ data_path = Path(dl_manager.download_and_extract(_URLS[_DATASETNAME]))
115
+
116
+ return [
117
+ datasets.SplitGenerator(
118
+ name=datasets.Split.TRAIN,
119
+ gen_kwargs={
120
+ "filepath": data_path,
121
+ "split": "train",
122
+ },
123
+ ),
124
+ ]
125
+
126
+ def _generate_examples(self, filepath: Path, split: str) -> Tuple[int, Dict]:
127
+ """Yields examples as (key, example) tuples."""
128
+
129
+ language = self.config.name.split("_")[3]
130
+
131
+ df = pd.read_parquet(filepath)
132
+ df = df[df["language"] == language]
133
+
134
+ for index, row in df.iterrows():
135
+ if self.config.schema == "source":
136
+ example = row.to_dict()
137
+
138
+ elif self.config.schema == "seacrowd_t2t":
139
+ example = {
140
+ "id": str(index),
141
+ "text_1": row["inputs"],
142
+ "text_2": row["targets"],
143
+ "text_1_name": "inputs",
144
+ "text_2_name": "targets",
145
+ }
146
+
147
+ yield index, example