Upload v1.py
Browse files
v1.py
ADDED
@@ -0,0 +1,154 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# -*- coding: utf-8 -*-
|
2 |
+
"""CLUTRR_Dataset Loading Script.ipynb
|
3 |
+
Automatically generated by Colaboratory.
|
4 |
+
Original file is located at
|
5 |
+
https://colab.research.google.com/drive/1q9DdeHA5JbgTHkH6kfZe_KWHQOwHZA97
|
6 |
+
"""
|
7 |
+
# coding=utf-8
|
8 |
+
# Copyright 2019 The CLUTRR Datasets Authors and the HuggingFace Datasets Authors.
|
9 |
+
#
|
10 |
+
# CLUTRR is CC-BY-NC 4.0 (Attr Non-Commercial Inter.) licensed, as found in the LICENSE file.
|
11 |
+
#
|
12 |
+
# Unless required by applicable law or agreed to in writing, software
|
13 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
14 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
15 |
+
# See the License for the specific language governing permissions and
|
16 |
+
# limitations under the License.
|
17 |
+
|
18 |
+
# Lint as: python3
|
19 |
+
"""The CLUTRR (Compositional Language Understanding and Text-based Relational Reasoning) benchmark."""
|
20 |
+
|
21 |
+
|
22 |
+
import csv
|
23 |
+
import os
|
24 |
+
import textwrap
|
25 |
+
|
26 |
+
import numpy as np
|
27 |
+
|
28 |
+
import datasets
|
29 |
+
import json
|
30 |
+
|
31 |
+
_CLUTRR_CITATION = """\
|
32 |
+
@article{sinha2019clutrr,
|
33 |
+
Author = {Koustuv Sinha and Shagun Sodhani and Jin Dong and Joelle Pineau and William L. Hamilton},
|
34 |
+
Title = {CLUTRR: A Diagnostic Benchmark for Inductive Reasoning from Text},
|
35 |
+
Year = {2019},
|
36 |
+
journal = {Empirical Methods of Natural Language Processing (EMNLP)},
|
37 |
+
arxiv = {1908.06177}
|
38 |
+
}
|
39 |
+
"""
|
40 |
+
|
41 |
+
_CLUTRR_DESCRIPTION = """\
|
42 |
+
CLUTRR (Compositional Language Understanding and Text-based Relational Reasoning),
|
43 |
+
a diagnostic benchmark suite, is first introduced in (https://arxiv.org/abs/1908.06177)
|
44 |
+
to test the systematic generalization and inductive reasoning capabilities of NLU systems.
|
45 |
+
"""
|
46 |
+
_URL = "https://raw.githubusercontent.com/kliang5/CLUTRR_huggingface_dataset/main/"
|
47 |
+
_TASK = ["gen_train23_test2to10", "gen_train234_test2to10", "rob_train_clean_23_test_all_23", "rob_train_disc_23_test_all_23", "rob_train_irr_23_test_all_23","rob_train_sup_23_test_all_23"]
|
48 |
+
|
49 |
+
class v1(datasets.GeneratorBasedBuilder):
|
50 |
+
"""BuilderConfig for CLUTRR."""
|
51 |
+
|
52 |
+
BUILDER_CONFIGS = [
|
53 |
+
datasets.BuilderConfig(
|
54 |
+
name=task,
|
55 |
+
version=datasets.Version("1.0.0"),
|
56 |
+
description="",
|
57 |
+
)
|
58 |
+
for task in _TASK
|
59 |
+
]
|
60 |
+
|
61 |
+
def _info(self):
|
62 |
+
return datasets.DatasetInfo(
|
63 |
+
description=_CLUTRR_DESCRIPTION,
|
64 |
+
features=datasets.Features(
|
65 |
+
{
|
66 |
+
"id": datasets.Value("string"),
|
67 |
+
"story": datasets.Value("string"),
|
68 |
+
"query": datasets.Value("string"),
|
69 |
+
"target": datasets.Value("int32"),
|
70 |
+
"target_text": datasets.Value("string"),
|
71 |
+
"clean_story": datasets.Value("string"),
|
72 |
+
"proof_state": datasets.Value("string"),
|
73 |
+
"f_comb": datasets.Value("string"),
|
74 |
+
"task_name": datasets.Value("string"),
|
75 |
+
"story_edges": datasets.Value("string"),
|
76 |
+
"edge_types": datasets.Value("string"),
|
77 |
+
"query_edge": datasets.Value("string"),
|
78 |
+
"genders": datasets.Value("string"),
|
79 |
+
"task_split": datasets.Value("string"),
|
80 |
+
}
|
81 |
+
),
|
82 |
+
# No default supervised_keys (as we have to pass both premise
|
83 |
+
# and hypothesis as input).
|
84 |
+
supervised_keys=None,
|
85 |
+
homepage="https://www.cs.mcgill.ca/~ksinha4/clutrr/",
|
86 |
+
citation=_CLUTRR_CITATION,
|
87 |
+
)
|
88 |
+
|
89 |
+
def _split_generators(self, dl_manager):
|
90 |
+
"""Returns SplitGenerators."""
|
91 |
+
# dl_manager is a datasets.download.DownloadManager that can be used to
|
92 |
+
# download and extract URLs
|
93 |
+
|
94 |
+
task = str(self.config.name)
|
95 |
+
urls_to_download = {
|
96 |
+
"test": _URL + task + "/test.csv",
|
97 |
+
"train": _URL + task + "/train.csv",
|
98 |
+
"validation": _URL + task + "/validation.csv",
|
99 |
+
}
|
100 |
+
downloaded_files = dl_manager.download_and_extract(urls_to_download)
|
101 |
+
|
102 |
+
|
103 |
+
return [
|
104 |
+
datasets.SplitGenerator(
|
105 |
+
name=datasets.Split.TRAIN,
|
106 |
+
# These kwargs will be passed to _generate_examples
|
107 |
+
gen_kwargs={
|
108 |
+
"filepath": downloaded_files["train"],
|
109 |
+
"task": task,
|
110 |
+
},
|
111 |
+
),
|
112 |
+
datasets.SplitGenerator(
|
113 |
+
name=datasets.Split.VALIDATION,
|
114 |
+
# These kwargs will be passed to _generate_examples
|
115 |
+
gen_kwargs={
|
116 |
+
"filepath": downloaded_files["validation"],
|
117 |
+
"task": task,
|
118 |
+
},
|
119 |
+
),
|
120 |
+
datasets.SplitGenerator(
|
121 |
+
name=datasets.Split.TEST,
|
122 |
+
# These kwargs will be passed to _generate_examples
|
123 |
+
gen_kwargs={
|
124 |
+
"filepath": downloaded_files["test"],
|
125 |
+
"task": task,
|
126 |
+
},
|
127 |
+
),
|
128 |
+
]
|
129 |
+
|
130 |
+
def _generate_examples(self, filepath, task):
|
131 |
+
"""Yields examples."""
|
132 |
+
with open(filepath, encoding="utf-8") as f:
|
133 |
+
reader = csv.reader(f)
|
134 |
+
for id_, data in enumerate(reader):
|
135 |
+
if id_ == 0:
|
136 |
+
continue
|
137 |
+
# yield id_, data
|
138 |
+
# id_ += 1
|
139 |
+
yield id_, {
|
140 |
+
"id": data[1],
|
141 |
+
"story": data[2],
|
142 |
+
"query": data[3],
|
143 |
+
"target": data[4],
|
144 |
+
"target_text": data[5],
|
145 |
+
"clean_story": data[6],
|
146 |
+
"proof_state": data[7],
|
147 |
+
"f_comb": data[8],
|
148 |
+
"task_name": data[9],
|
149 |
+
"story_edges": data[10],
|
150 |
+
"edge_types": data[11],
|
151 |
+
"query_edge": data[12],
|
152 |
+
"genders": data[13],
|
153 |
+
"task_split": data[14],
|
154 |
+
}
|