Commit
•
14dd78b
1
Parent(s):
83737e5
Delete loading script
Browse files
cfq.py
DELETED
@@ -1,167 +0,0 @@
|
|
1 |
-
# coding=utf-8
|
2 |
-
# Copyright 2020 The TensorFlow Datasets Authors and the HuggingFace Datasets Authors.
|
3 |
-
#
|
4 |
-
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
-
# you may not use this file except in compliance with the License.
|
6 |
-
# You may obtain a copy of the License at
|
7 |
-
#
|
8 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
-
#
|
10 |
-
# Unless required by applicable law or agreed to in writing, software
|
11 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
-
# See the License for the specific language governing permissions and
|
14 |
-
# limitations under the License.
|
15 |
-
|
16 |
-
# Lint as: python3
|
17 |
-
"""CFQ (Compositional Freebase Questions) dataset."""
|
18 |
-
|
19 |
-
|
20 |
-
import json
|
21 |
-
import re
|
22 |
-
|
23 |
-
import datasets
|
24 |
-
|
25 |
-
|
26 |
-
logger = datasets.logging.get_logger(__name__)
|
27 |
-
|
28 |
-
|
29 |
-
_HOMEPAGE = "https://github.com/google-research/google-research/tree/master/cfq"
|
30 |
-
|
31 |
-
_LICENSE = "CC BY 4.0"
|
32 |
-
|
33 |
-
_CITATION = """
|
34 |
-
@inproceedings{Keysers2020,
|
35 |
-
title={Measuring Compositional Generalization: A Comprehensive Method on
|
36 |
-
Realistic Data},
|
37 |
-
author={Daniel Keysers and Nathanael Sch\"{a}rli and Nathan Scales and
|
38 |
-
Hylke Buisman and Daniel Furrer and Sergii Kashubin and
|
39 |
-
Nikola Momchev and Danila Sinopalnikov and Lukasz Stafiniak and
|
40 |
-
Tibor Tihon and Dmitry Tsarkov and Xiao Wang and Marc van Zee and
|
41 |
-
Olivier Bousquet},
|
42 |
-
booktitle={ICLR},
|
43 |
-
year={2020},
|
44 |
-
url={https://arxiv.org/abs/1912.09713.pdf},
|
45 |
-
}
|
46 |
-
"""
|
47 |
-
|
48 |
-
_DESCRIPTION = """
|
49 |
-
The CFQ dataset (and it's splits) for measuring compositional generalization.
|
50 |
-
|
51 |
-
See https://arxiv.org/abs/1912.09713.pdf for background.
|
52 |
-
|
53 |
-
Example usage:
|
54 |
-
data = datasets.load_dataset('cfq/mcd1')
|
55 |
-
"""
|
56 |
-
|
57 |
-
_DATA_URL = "https://storage.googleapis.com/cfq_dataset/cfq.tar.gz"
|
58 |
-
|
59 |
-
|
60 |
-
class CfqConfig(datasets.BuilderConfig):
|
61 |
-
"""BuilderConfig for CFQ splits."""
|
62 |
-
|
63 |
-
def __init__(self, name, directory="splits", **kwargs):
|
64 |
-
"""BuilderConfig for CFQ.
|
65 |
-
|
66 |
-
Args:
|
67 |
-
name: Unique name of the split.
|
68 |
-
directory: Which subdirectory to read the split from.
|
69 |
-
**kwargs: keyword arguments forwarded to super.
|
70 |
-
"""
|
71 |
-
# Version history:
|
72 |
-
super(CfqConfig, self).__init__(
|
73 |
-
name=name, version=datasets.Version("1.0.1"), description=_DESCRIPTION, **kwargs
|
74 |
-
)
|
75 |
-
self.splits_path = f"cfq/{directory}/{name}.json"
|
76 |
-
|
77 |
-
|
78 |
-
_QUESTION = "question"
|
79 |
-
_QUERY = "query"
|
80 |
-
_QUESTION_FIELD = "questionPatternModEntities"
|
81 |
-
_QUERY_FIELD = "sparqlPatternModEntities"
|
82 |
-
|
83 |
-
|
84 |
-
class Cfq(datasets.GeneratorBasedBuilder):
|
85 |
-
"""CFQ task / splits."""
|
86 |
-
|
87 |
-
BUILDER_CONFIGS = [
|
88 |
-
CfqConfig(name="mcd1"),
|
89 |
-
CfqConfig(name="mcd2"),
|
90 |
-
CfqConfig(name="mcd3"),
|
91 |
-
CfqConfig(name="question_complexity_split"),
|
92 |
-
CfqConfig(name="question_pattern_split"),
|
93 |
-
CfqConfig(name="query_complexity_split"),
|
94 |
-
CfqConfig(name="query_pattern_split"),
|
95 |
-
CfqConfig(name="random_split"),
|
96 |
-
]
|
97 |
-
|
98 |
-
def _info(self):
|
99 |
-
return datasets.DatasetInfo(
|
100 |
-
description=_DESCRIPTION,
|
101 |
-
features=datasets.Features(
|
102 |
-
{
|
103 |
-
_QUESTION: datasets.Value("string"),
|
104 |
-
_QUERY: datasets.Value("string"),
|
105 |
-
}
|
106 |
-
),
|
107 |
-
supervised_keys=(_QUESTION, _QUERY),
|
108 |
-
homepage=_HOMEPAGE,
|
109 |
-
license=_LICENSE,
|
110 |
-
citation=_CITATION,
|
111 |
-
)
|
112 |
-
|
113 |
-
def _split_generators(self, dl_manager):
|
114 |
-
"""Returns SplitGenerators."""
|
115 |
-
archive_path = dl_manager.download(_DATA_URL)
|
116 |
-
return [
|
117 |
-
datasets.SplitGenerator(
|
118 |
-
name=split,
|
119 |
-
gen_kwargs={
|
120 |
-
"data_files": dl_manager.iter_archive(archive_path),
|
121 |
-
"split_id": f"{split}Idxs",
|
122 |
-
},
|
123 |
-
)
|
124 |
-
for split in [datasets.Split.TRAIN, datasets.Split.TEST]
|
125 |
-
]
|
126 |
-
|
127 |
-
def _scrub_json(self, content):
|
128 |
-
"""Reduce JSON by filtering out only the fields of interest."""
|
129 |
-
# Loading of json data with the standard Python library is very inefficient:
|
130 |
-
# For the 4GB dataset file it requires more than 40GB of RAM and takes 3min.
|
131 |
-
# There are more efficient libraries but in order to avoid additional
|
132 |
-
# dependencies we use a simple (perhaps somewhat brittle) regexp to reduce
|
133 |
-
# the content to only what is needed.
|
134 |
-
question_regex = re.compile(r'("%s":\s*"[^"]*")' % _QUESTION_FIELD)
|
135 |
-
query_regex = re.compile(r'("%s":\s*"[^"]*")' % _QUERY_FIELD)
|
136 |
-
question_match = None
|
137 |
-
for line in content:
|
138 |
-
line = line.decode("utf-8")
|
139 |
-
if not question_match:
|
140 |
-
question_match = question_regex.match(line)
|
141 |
-
else:
|
142 |
-
query_match = query_regex.match(line)
|
143 |
-
if query_match:
|
144 |
-
yield json.loads("{" + question_match.group(1) + "," + query_match.group(1) + "}")
|
145 |
-
question_match = None
|
146 |
-
|
147 |
-
def _generate_examples(self, data_files, split_id):
|
148 |
-
"""Yields examples."""
|
149 |
-
samples_path = "cfq/dataset.json"
|
150 |
-
for path, file in data_files:
|
151 |
-
if path == self.config.splits_path:
|
152 |
-
splits = json.load(file)[split_id]
|
153 |
-
elif path == samples_path:
|
154 |
-
# The samples_path is the last path inside the archive
|
155 |
-
generator = enumerate(self._scrub_json(file))
|
156 |
-
samples = {}
|
157 |
-
splits_set = set(splits)
|
158 |
-
for split_idx in splits:
|
159 |
-
if split_idx in samples:
|
160 |
-
sample = samples.pop(split_idx)
|
161 |
-
else:
|
162 |
-
for sample_idx, sample in generator:
|
163 |
-
if sample_idx == split_idx:
|
164 |
-
break
|
165 |
-
elif sample_idx in splits_set:
|
166 |
-
samples[sample_idx] = sample
|
167 |
-
yield split_idx, {_QUESTION: sample[_QUESTION_FIELD], _QUERY: sample[_QUERY_FIELD]}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|