adds first dataloader
Browse files
art.py
ADDED
@@ -0,0 +1,84 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import json
|
2 |
+
import os
|
3 |
+
|
4 |
+
import datasets
|
5 |
+
|
6 |
+
|
7 |
+
_CITATION = """\
|
8 |
+
@InProceedings{anli,
|
9 |
+
author = {Chandra, Bhagavatula and Ronan, Le Bras and Chaitanya, Malaviya and Keisuke, Sakaguchi and Ari, Holtzman
|
10 |
+
and Hannah, Rashkin and Doug, Downey and Scott, Wen-tau Yih and Yejin, Choi},
|
11 |
+
title = {Abductive Commonsense Reasoning},
|
12 |
+
year = {2020}
|
13 |
+
}"""
|
14 |
+
|
15 |
+
_DESCRIPTION = """\
|
16 |
+
the Abductive Natural Language Generation Dataset from AI2
|
17 |
+
"""
|
18 |
+
_DATA_URL = "https://storage.googleapis.com/ai2-mosaic/public/abductive-commonsense-reasoning-iclr2020/anlg.zip"
|
19 |
+
_HOMEPAGE = "https://github.com/allenai/abductive-commonsense-reasoning"
|
20 |
+
|
21 |
+
class ArtConfig(datasets.BuilderConfig):
|
22 |
+
"""BuilderConfig for Art."""
|
23 |
+
|
24 |
+
def __init__(self, **kwargs):
|
25 |
+
"""BuilderConfig for Art.
|
26 |
+
Args:
|
27 |
+
**kwargs: keyword arguments forwarded to super.
|
28 |
+
"""
|
29 |
+
super(ArtConfig, self).__init__(version=datasets.Version("0.1.0", ""), **kwargs)
|
30 |
+
|
31 |
+
|
32 |
+
class Art(datasets.GeneratorBasedBuilder):
|
33 |
+
VERSION = datasets.Version("0.1.0")
|
34 |
+
BUILDER_CONFIGS = [
|
35 |
+
ArtConfig(
|
36 |
+
name="anlg",
|
37 |
+
description="""\
|
38 |
+
Abductive Natural Language Generation Dataset from AI2.
|
39 |
+
""",
|
40 |
+
),
|
41 |
+
]
|
42 |
+
|
43 |
+
def _info(self):
|
44 |
+
return datasets.DatasetInfo(
|
45 |
+
description=_DESCRIPTION,
|
46 |
+
features=datasets.Features(
|
47 |
+
{
|
48 |
+
"gem_id": datasets.Value("string"),
|
49 |
+
"observation_1": datasets.Value("string"),
|
50 |
+
"observation_2": datasets.Value("string"),
|
51 |
+
"label": datasets.Value("string"),
|
52 |
+
}
|
53 |
+
),
|
54 |
+
homepage=_HOMEPAGE,
|
55 |
+
citation=_CITATION,
|
56 |
+
)
|
57 |
+
|
58 |
+
def _split_generators(self, dl_manager):
|
59 |
+
ds_splits = [datasets.Split.TRAIN, datasets.Split.VALIDATION, datasets.Split.TEST]
|
60 |
+
splits = ["train", "dev", "test"]
|
61 |
+
dl_dir = dl_manager.download_and_extract(_DATA_URL)
|
62 |
+
|
63 |
+
return [
|
64 |
+
datasets.SplitGenerator(
|
65 |
+
name=ds_split,
|
66 |
+
gen_kwargs={
|
67 |
+
"filepath": os.path.join(dl_dir, "anlg", f"{split}-w-comet-preds.jsonl"),
|
68 |
+
"split": split if split != "dev" else "validation" # adheres to GEM naming conventions
|
69 |
+
},
|
70 |
+
) for ds_split, split in zip(ds_splits, splits)
|
71 |
+
]
|
72 |
+
|
73 |
+
def _generate_examples(self, filepath, split):
|
74 |
+
with open(filepath, "r", encoding="utf-8") as f:
|
75 |
+
data = [json.loads(line) for line in f.readlines()]
|
76 |
+
|
77 |
+
for idx, row in enumerate(data):
|
78 |
+
label = row[f"hyp{row['label']}"]
|
79 |
+
yield idx, {
|
80 |
+
"gem_id": f"GEM-ART-{split}-{idx}",
|
81 |
+
"observation_1": row["obs1"],
|
82 |
+
"observation_2": row["obs2"],
|
83 |
+
"label": label,
|
84 |
+
}
|