ianporada parquet-converter commited on
Commit
501fbb8
0 Parent(s):

Duplicate from winogrande

Browse files

Co-authored-by: Parquet-converter (BOT) <parquet-converter@users.noreply.huggingface.co>

Files changed (4) hide show
  1. .gitattributes +27 -0
  2. README.md +365 -0
  3. dataset_infos.json +1 -0
  4. winogrande.py +145 -0
.gitattributes ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bin.* filter=lfs diff=lfs merge=lfs -text
5
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.model filter=lfs diff=lfs merge=lfs -text
12
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
13
+ *.onnx filter=lfs diff=lfs merge=lfs -text
14
+ *.ot filter=lfs diff=lfs merge=lfs -text
15
+ *.parquet filter=lfs diff=lfs merge=lfs -text
16
+ *.pb filter=lfs diff=lfs merge=lfs -text
17
+ *.pt filter=lfs diff=lfs merge=lfs -text
18
+ *.pth filter=lfs diff=lfs merge=lfs -text
19
+ *.rar filter=lfs diff=lfs merge=lfs -text
20
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
21
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
22
+ *.tflite filter=lfs diff=lfs merge=lfs -text
23
+ *.tgz filter=lfs diff=lfs merge=lfs -text
24
+ *.xz filter=lfs diff=lfs merge=lfs -text
25
+ *.zip filter=lfs diff=lfs merge=lfs -text
26
+ *.zstandard filter=lfs diff=lfs merge=lfs -text
27
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,365 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ paperswithcode_id: winogrande
5
+ pretty_name: WinoGrande
6
+ dataset_info:
7
+ - config_name: winogrande_xs
8
+ features:
9
+ - name: sentence
10
+ dtype: string
11
+ - name: option1
12
+ dtype: string
13
+ - name: option2
14
+ dtype: string
15
+ - name: answer
16
+ dtype: string
17
+ splits:
18
+ - name: train
19
+ num_bytes: 20704
20
+ num_examples: 160
21
+ - name: test
22
+ num_bytes: 227649
23
+ num_examples: 1767
24
+ - name: validation
25
+ num_bytes: 164199
26
+ num_examples: 1267
27
+ download_size: 3395492
28
+ dataset_size: 412552
29
+ - config_name: winogrande_s
30
+ features:
31
+ - name: sentence
32
+ dtype: string
33
+ - name: option1
34
+ dtype: string
35
+ - name: option2
36
+ dtype: string
37
+ - name: answer
38
+ dtype: string
39
+ splits:
40
+ - name: train
41
+ num_bytes: 82308
42
+ num_examples: 640
43
+ - name: test
44
+ num_bytes: 227649
45
+ num_examples: 1767
46
+ - name: validation
47
+ num_bytes: 164199
48
+ num_examples: 1267
49
+ download_size: 3395492
50
+ dataset_size: 474156
51
+ - config_name: winogrande_m
52
+ features:
53
+ - name: sentence
54
+ dtype: string
55
+ - name: option1
56
+ dtype: string
57
+ - name: option2
58
+ dtype: string
59
+ - name: answer
60
+ dtype: string
61
+ splits:
62
+ - name: train
63
+ num_bytes: 329001
64
+ num_examples: 2558
65
+ - name: test
66
+ num_bytes: 227649
67
+ num_examples: 1767
68
+ - name: validation
69
+ num_bytes: 164199
70
+ num_examples: 1267
71
+ download_size: 3395492
72
+ dataset_size: 720849
73
+ - config_name: winogrande_l
74
+ features:
75
+ - name: sentence
76
+ dtype: string
77
+ - name: option1
78
+ dtype: string
79
+ - name: option2
80
+ dtype: string
81
+ - name: answer
82
+ dtype: string
83
+ splits:
84
+ - name: train
85
+ num_bytes: 1319576
86
+ num_examples: 10234
87
+ - name: test
88
+ num_bytes: 227649
89
+ num_examples: 1767
90
+ - name: validation
91
+ num_bytes: 164199
92
+ num_examples: 1267
93
+ download_size: 3395492
94
+ dataset_size: 1711424
95
+ - config_name: winogrande_xl
96
+ features:
97
+ - name: sentence
98
+ dtype: string
99
+ - name: option1
100
+ dtype: string
101
+ - name: option2
102
+ dtype: string
103
+ - name: answer
104
+ dtype: string
105
+ splits:
106
+ - name: train
107
+ num_bytes: 5185832
108
+ num_examples: 40398
109
+ - name: test
110
+ num_bytes: 227649
111
+ num_examples: 1767
112
+ - name: validation
113
+ num_bytes: 164199
114
+ num_examples: 1267
115
+ download_size: 3395492
116
+ dataset_size: 5577680
117
+ - config_name: winogrande_debiased
118
+ features:
119
+ - name: sentence
120
+ dtype: string
121
+ - name: option1
122
+ dtype: string
123
+ - name: option2
124
+ dtype: string
125
+ - name: answer
126
+ dtype: string
127
+ splits:
128
+ - name: train
129
+ num_bytes: 1203420
130
+ num_examples: 9248
131
+ - name: test
132
+ num_bytes: 227649
133
+ num_examples: 1767
134
+ - name: validation
135
+ num_bytes: 164199
136
+ num_examples: 1267
137
+ download_size: 3395492
138
+ dataset_size: 1595268
139
+ ---
140
+
141
+ # Dataset Card for "winogrande"
142
+
143
+ ## Table of Contents
144
+ - [Dataset Description](#dataset-description)
145
+ - [Dataset Summary](#dataset-summary)
146
+ - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
147
+ - [Languages](#languages)
148
+ - [Dataset Structure](#dataset-structure)
149
+ - [Data Instances](#data-instances)
150
+ - [Data Fields](#data-fields)
151
+ - [Data Splits](#data-splits)
152
+ - [Dataset Creation](#dataset-creation)
153
+ - [Curation Rationale](#curation-rationale)
154
+ - [Source Data](#source-data)
155
+ - [Annotations](#annotations)
156
+ - [Personal and Sensitive Information](#personal-and-sensitive-information)
157
+ - [Considerations for Using the Data](#considerations-for-using-the-data)
158
+ - [Social Impact of Dataset](#social-impact-of-dataset)
159
+ - [Discussion of Biases](#discussion-of-biases)
160
+ - [Other Known Limitations](#other-known-limitations)
161
+ - [Additional Information](#additional-information)
162
+ - [Dataset Curators](#dataset-curators)
163
+ - [Licensing Information](#licensing-information)
164
+ - [Citation Information](#citation-information)
165
+ - [Contributions](#contributions)
166
+
167
+ ## Dataset Description
168
+
169
+ - **Homepage:** [https://leaderboard.allenai.org/winogrande/submissions/get-started](https://leaderboard.allenai.org/winogrande/submissions/get-started)
170
+ - **Repository:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
171
+ - **Paper:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
172
+ - **Point of Contact:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
173
+ - **Size of downloaded dataset files:** 20.37 MB
174
+ - **Size of the generated dataset:** 10.50 MB
175
+ - **Total amount of disk used:** 30.87 MB
176
+
177
+ ### Dataset Summary
178
+
179
+ WinoGrande is a new collection of 44k problems, inspired by Winograd Schema Challenge (Levesque, Davis, and Morgenstern
180
+ 2011), but adjusted to improve the scale and robustness against the dataset-specific bias. Formulated as a
181
+ fill-in-a-blank task with binary options, the goal is to choose the right option for a given sentence which requires
182
+ commonsense reasoning.
183
+
184
+ ### Supported Tasks and Leaderboards
185
+
186
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
187
+
188
+ ### Languages
189
+
190
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
191
+
192
+ ## Dataset Structure
193
+
194
+ ### Data Instances
195
+
196
+ #### winogrande_debiased
197
+
198
+ - **Size of downloaded dataset files:** 3.40 MB
199
+ - **Size of the generated dataset:** 1.59 MB
200
+ - **Total amount of disk used:** 4.99 MB
201
+
202
+ An example of 'train' looks as follows.
203
+ ```
204
+
205
+ ```
206
+
207
+ #### winogrande_l
208
+
209
+ - **Size of downloaded dataset files:** 3.40 MB
210
+ - **Size of the generated dataset:** 1.71 MB
211
+ - **Total amount of disk used:** 5.11 MB
212
+
213
+ An example of 'validation' looks as follows.
214
+ ```
215
+
216
+ ```
217
+
218
+ #### winogrande_m
219
+
220
+ - **Size of downloaded dataset files:** 3.40 MB
221
+ - **Size of the generated dataset:** 0.72 MB
222
+ - **Total amount of disk used:** 4.12 MB
223
+
224
+ An example of 'validation' looks as follows.
225
+ ```
226
+
227
+ ```
228
+
229
+ #### winogrande_s
230
+
231
+ - **Size of downloaded dataset files:** 3.40 MB
232
+ - **Size of the generated dataset:** 0.47 MB
233
+ - **Total amount of disk used:** 3.87 MB
234
+
235
+ An example of 'validation' looks as follows.
236
+ ```
237
+
238
+ ```
239
+
240
+ #### winogrande_xl
241
+
242
+ - **Size of downloaded dataset files:** 3.40 MB
243
+ - **Size of the generated dataset:** 5.58 MB
244
+ - **Total amount of disk used:** 8.98 MB
245
+
246
+ An example of 'train' looks as follows.
247
+ ```
248
+
249
+ ```
250
+
251
+ ### Data Fields
252
+
253
+ The data fields are the same among all splits.
254
+
255
+ #### winogrande_debiased
256
+ - `sentence`: a `string` feature.
257
+ - `option1`: a `string` feature.
258
+ - `option2`: a `string` feature.
259
+ - `answer`: a `string` feature.
260
+
261
+ #### winogrande_l
262
+ - `sentence`: a `string` feature.
263
+ - `option1`: a `string` feature.
264
+ - `option2`: a `string` feature.
265
+ - `answer`: a `string` feature.
266
+
267
+ #### winogrande_m
268
+ - `sentence`: a `string` feature.
269
+ - `option1`: a `string` feature.
270
+ - `option2`: a `string` feature.
271
+ - `answer`: a `string` feature.
272
+
273
+ #### winogrande_s
274
+ - `sentence`: a `string` feature.
275
+ - `option1`: a `string` feature.
276
+ - `option2`: a `string` feature.
277
+ - `answer`: a `string` feature.
278
+
279
+ #### winogrande_xl
280
+ - `sentence`: a `string` feature.
281
+ - `option1`: a `string` feature.
282
+ - `option2`: a `string` feature.
283
+ - `answer`: a `string` feature.
284
+
285
+ ### Data Splits
286
+
287
+ | name |train|validation|test|
288
+ |-------------------|----:|---------:|---:|
289
+ |winogrande_debiased| 9248| 1267|1767|
290
+ |winogrande_l |10234| 1267|1767|
291
+ |winogrande_m | 2558| 1267|1767|
292
+ |winogrande_s | 640| 1267|1767|
293
+ |winogrande_xl |40398| 1267|1767|
294
+ |winogrande_xs | 160| 1267|1767|
295
+
296
+ ## Dataset Creation
297
+
298
+ ### Curation Rationale
299
+
300
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
301
+
302
+ ### Source Data
303
+
304
+ #### Initial Data Collection and Normalization
305
+
306
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
307
+
308
+ #### Who are the source language producers?
309
+
310
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
311
+
312
+ ### Annotations
313
+
314
+ #### Annotation process
315
+
316
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
317
+
318
+ #### Who are the annotators?
319
+
320
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
321
+
322
+ ### Personal and Sensitive Information
323
+
324
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
325
+
326
+ ## Considerations for Using the Data
327
+
328
+ ### Social Impact of Dataset
329
+
330
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
331
+
332
+ ### Discussion of Biases
333
+
334
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
335
+
336
+ ### Other Known Limitations
337
+
338
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
339
+
340
+ ## Additional Information
341
+
342
+ ### Dataset Curators
343
+
344
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
345
+
346
+ ### Licensing Information
347
+
348
+ [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
349
+
350
+ ### Citation Information
351
+
352
+ ```
353
+ @InProceedings{ai2:winogrande,
354
+ title = {WinoGrande: An Adversarial Winograd Schema Challenge at Scale},
355
+ authors={Keisuke, Sakaguchi and Ronan, Le Bras and Chandra, Bhagavatula and Yejin, Choi
356
+ },
357
+ year={2019}
358
+ }
359
+
360
+ ```
361
+
362
+
363
+ ### Contributions
364
+
365
+ Thanks to [@thomwolf](https://github.com/thomwolf), [@TevenLeScao](https://github.com/TevenLeScao), [@patrickvonplaten](https://github.com/patrickvonplaten), [@lewtun](https://github.com/lewtun) for adding this dataset.
dataset_infos.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"winogrande_xs": {"description": "WinoGrande is a new collection of 44k problems, inspired by Winograd Schema Challenge (Levesque, Davis, and Morgenstern\n 2011), but adjusted to improve the scale and robustness against the dataset-specific bias. Formulated as a\nfill-in-a-blank task with binary options, the goal is to choose the right option for a given sentence which requires\ncommonsense reasoning.\n", "citation": "@InProceedings{ai2:winogrande,\ntitle = {WinoGrande: An Adversarial Winograd Schema Challenge at Scale},\nauthors={Keisuke, Sakaguchi and Ronan, Le Bras and Chandra, Bhagavatula and Yejin, Choi\n},\nyear={2019}\n}\n", "homepage": "https://leaderboard.allenai.org/winogrande/submissions/get-started", "license": "", "features": {"sentence": {"dtype": "string", "id": null, "_type": "Value"}, "option1": {"dtype": "string", "id": null, "_type": "Value"}, "option2": {"dtype": "string", "id": null, "_type": "Value"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "winogrande", "config_name": "winogrande_xs", "version": {"version_str": "1.1.0", "description": "", "major": 1, "minor": 1, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 20704, "num_examples": 160, "dataset_name": "winogrande"}, "test": {"name": "test", "num_bytes": 227649, "num_examples": 1767, "dataset_name": "winogrande"}, "validation": {"name": "validation", "num_bytes": 164199, "num_examples": 1267, "dataset_name": "winogrande"}}, "download_checksums": {"https://storage.googleapis.com/ai2-mosaic/public/winogrande/winogrande_1.1.zip": {"num_bytes": 3395492, "checksum": "3619ab104d8be2977b25c90ff420cb42d491707dcc75362a1e5d22bc082b7318"}}, "download_size": 3395492, "post_processing_size": null, "dataset_size": 412552, "size_in_bytes": 3808044}, "winogrande_s": {"description": "WinoGrande is a new collection of 44k problems, inspired by Winograd Schema Challenge (Levesque, Davis, and Morgenstern\n 2011), but adjusted to improve the scale and robustness against the dataset-specific bias. Formulated as a\nfill-in-a-blank task with binary options, the goal is to choose the right option for a given sentence which requires\ncommonsense reasoning.\n", "citation": "@InProceedings{ai2:winogrande,\ntitle = {WinoGrande: An Adversarial Winograd Schema Challenge at Scale},\nauthors={Keisuke, Sakaguchi and Ronan, Le Bras and Chandra, Bhagavatula and Yejin, Choi\n},\nyear={2019}\n}\n", "homepage": "https://leaderboard.allenai.org/winogrande/submissions/get-started", "license": "", "features": {"sentence": {"dtype": "string", "id": null, "_type": "Value"}, "option1": {"dtype": "string", "id": null, "_type": "Value"}, "option2": {"dtype": "string", "id": null, "_type": "Value"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "winogrande", "config_name": "winogrande_s", "version": {"version_str": "1.1.0", "description": "", "major": 1, "minor": 1, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 82308, "num_examples": 640, "dataset_name": "winogrande"}, "test": {"name": "test", "num_bytes": 227649, "num_examples": 1767, "dataset_name": "winogrande"}, "validation": {"name": "validation", "num_bytes": 164199, "num_examples": 1267, "dataset_name": "winogrande"}}, "download_checksums": {"https://storage.googleapis.com/ai2-mosaic/public/winogrande/winogrande_1.1.zip": {"num_bytes": 3395492, "checksum": "3619ab104d8be2977b25c90ff420cb42d491707dcc75362a1e5d22bc082b7318"}}, "download_size": 3395492, "post_processing_size": null, "dataset_size": 474156, "size_in_bytes": 3869648}, "winogrande_m": {"description": "WinoGrande is a new collection of 44k problems, inspired by Winograd Schema Challenge (Levesque, Davis, and Morgenstern\n 2011), but adjusted to improve the scale and robustness against the dataset-specific bias. Formulated as a\nfill-in-a-blank task with binary options, the goal is to choose the right option for a given sentence which requires\ncommonsense reasoning.\n", "citation": "@InProceedings{ai2:winogrande,\ntitle = {WinoGrande: An Adversarial Winograd Schema Challenge at Scale},\nauthors={Keisuke, Sakaguchi and Ronan, Le Bras and Chandra, Bhagavatula and Yejin, Choi\n},\nyear={2019}\n}\n", "homepage": "https://leaderboard.allenai.org/winogrande/submissions/get-started", "license": "", "features": {"sentence": {"dtype": "string", "id": null, "_type": "Value"}, "option1": {"dtype": "string", "id": null, "_type": "Value"}, "option2": {"dtype": "string", "id": null, "_type": "Value"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "winogrande", "config_name": "winogrande_m", "version": {"version_str": "1.1.0", "description": "", "major": 1, "minor": 1, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 329001, "num_examples": 2558, "dataset_name": "winogrande"}, "test": {"name": "test", "num_bytes": 227649, "num_examples": 1767, "dataset_name": "winogrande"}, "validation": {"name": "validation", "num_bytes": 164199, "num_examples": 1267, "dataset_name": "winogrande"}}, "download_checksums": {"https://storage.googleapis.com/ai2-mosaic/public/winogrande/winogrande_1.1.zip": {"num_bytes": 3395492, "checksum": "3619ab104d8be2977b25c90ff420cb42d491707dcc75362a1e5d22bc082b7318"}}, "download_size": 3395492, "post_processing_size": null, "dataset_size": 720849, "size_in_bytes": 4116341}, "winogrande_l": {"description": "WinoGrande is a new collection of 44k problems, inspired by Winograd Schema Challenge (Levesque, Davis, and Morgenstern\n 2011), but adjusted to improve the scale and robustness against the dataset-specific bias. Formulated as a\nfill-in-a-blank task with binary options, the goal is to choose the right option for a given sentence which requires\ncommonsense reasoning.\n", "citation": "@InProceedings{ai2:winogrande,\ntitle = {WinoGrande: An Adversarial Winograd Schema Challenge at Scale},\nauthors={Keisuke, Sakaguchi and Ronan, Le Bras and Chandra, Bhagavatula and Yejin, Choi\n},\nyear={2019}\n}\n", "homepage": "https://leaderboard.allenai.org/winogrande/submissions/get-started", "license": "", "features": {"sentence": {"dtype": "string", "id": null, "_type": "Value"}, "option1": {"dtype": "string", "id": null, "_type": "Value"}, "option2": {"dtype": "string", "id": null, "_type": "Value"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "winogrande", "config_name": "winogrande_l", "version": {"version_str": "1.1.0", "description": "", "major": 1, "minor": 1, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 1319576, "num_examples": 10234, "dataset_name": "winogrande"}, "test": {"name": "test", "num_bytes": 227649, "num_examples": 1767, "dataset_name": "winogrande"}, "validation": {"name": "validation", "num_bytes": 164199, "num_examples": 1267, "dataset_name": "winogrande"}}, "download_checksums": {"https://storage.googleapis.com/ai2-mosaic/public/winogrande/winogrande_1.1.zip": {"num_bytes": 3395492, "checksum": "3619ab104d8be2977b25c90ff420cb42d491707dcc75362a1e5d22bc082b7318"}}, "download_size": 3395492, "post_processing_size": null, "dataset_size": 1711424, "size_in_bytes": 5106916}, "winogrande_xl": {"description": "WinoGrande is a new collection of 44k problems, inspired by Winograd Schema Challenge (Levesque, Davis, and Morgenstern\n 2011), but adjusted to improve the scale and robustness against the dataset-specific bias. Formulated as a\nfill-in-a-blank task with binary options, the goal is to choose the right option for a given sentence which requires\ncommonsense reasoning.\n", "citation": "@InProceedings{ai2:winogrande,\ntitle = {WinoGrande: An Adversarial Winograd Schema Challenge at Scale},\nauthors={Keisuke, Sakaguchi and Ronan, Le Bras and Chandra, Bhagavatula and Yejin, Choi\n},\nyear={2019}\n}\n", "homepage": "https://leaderboard.allenai.org/winogrande/submissions/get-started", "license": "", "features": {"sentence": {"dtype": "string", "id": null, "_type": "Value"}, "option1": {"dtype": "string", "id": null, "_type": "Value"}, "option2": {"dtype": "string", "id": null, "_type": "Value"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "winogrande", "config_name": "winogrande_xl", "version": {"version_str": "1.1.0", "description": "", "major": 1, "minor": 1, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 5185832, "num_examples": 40398, "dataset_name": "winogrande"}, "test": {"name": "test", "num_bytes": 227649, "num_examples": 1767, "dataset_name": "winogrande"}, "validation": {"name": "validation", "num_bytes": 164199, "num_examples": 1267, "dataset_name": "winogrande"}}, "download_checksums": {"https://storage.googleapis.com/ai2-mosaic/public/winogrande/winogrande_1.1.zip": {"num_bytes": 3395492, "checksum": "3619ab104d8be2977b25c90ff420cb42d491707dcc75362a1e5d22bc082b7318"}}, "download_size": 3395492, "post_processing_size": null, "dataset_size": 5577680, "size_in_bytes": 8973172}, "winogrande_debiased": {"description": "WinoGrande is a new collection of 44k problems, inspired by Winograd Schema Challenge (Levesque, Davis, and Morgenstern\n 2011), but adjusted to improve the scale and robustness against the dataset-specific bias. Formulated as a\nfill-in-a-blank task with binary options, the goal is to choose the right option for a given sentence which requires\ncommonsense reasoning.\n", "citation": "@InProceedings{ai2:winogrande,\ntitle = {WinoGrande: An Adversarial Winograd Schema Challenge at Scale},\nauthors={Keisuke, Sakaguchi and Ronan, Le Bras and Chandra, Bhagavatula and Yejin, Choi\n},\nyear={2019}\n}\n", "homepage": "https://leaderboard.allenai.org/winogrande/submissions/get-started", "license": "", "features": {"sentence": {"dtype": "string", "id": null, "_type": "Value"}, "option1": {"dtype": "string", "id": null, "_type": "Value"}, "option2": {"dtype": "string", "id": null, "_type": "Value"}, "answer": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "builder_name": "winogrande", "config_name": "winogrande_debiased", "version": {"version_str": "1.1.0", "description": "", "major": 1, "minor": 1, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 1203420, "num_examples": 9248, "dataset_name": "winogrande"}, "test": {"name": "test", "num_bytes": 227649, "num_examples": 1767, "dataset_name": "winogrande"}, "validation": {"name": "validation", "num_bytes": 164199, "num_examples": 1267, "dataset_name": "winogrande"}}, "download_checksums": {"https://storage.googleapis.com/ai2-mosaic/public/winogrande/winogrande_1.1.zip": {"num_bytes": 3395492, "checksum": "3619ab104d8be2977b25c90ff420cb42d491707dcc75362a1e5d22bc082b7318"}}, "download_size": 3395492, "post_processing_size": null, "dataset_size": 1595268, "size_in_bytes": 4990760}}
winogrande.py ADDED
@@ -0,0 +1,145 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """TODO(winogrande): Add a description here."""
2
+
3
+
4
+ import json
5
+ import os
6
+
7
+ import datasets
8
+
9
+
10
+ # TODO(winogrande): BibTeX citation
11
+ _CITATION = """\
12
+ @InProceedings{ai2:winogrande,
13
+ title = {WinoGrande: An Adversarial Winograd Schema Challenge at Scale},
14
+ authors={Keisuke, Sakaguchi and Ronan, Le Bras and Chandra, Bhagavatula and Yejin, Choi
15
+ },
16
+ year={2019}
17
+ }
18
+ """
19
+
20
+ # TODO(winogrande):
21
+ _DESCRIPTION = """\
22
+ WinoGrande is a new collection of 44k problems, inspired by Winograd Schema Challenge (Levesque, Davis, and Morgenstern
23
+ 2011), but adjusted to improve the scale and robustness against the dataset-specific bias. Formulated as a
24
+ fill-in-a-blank task with binary options, the goal is to choose the right option for a given sentence which requires
25
+ commonsense reasoning.
26
+ """
27
+
28
+ _URL = "https://storage.googleapis.com/ai2-mosaic/public/winogrande/winogrande_1.1.zip"
29
+ _FORMATS = ["xs", "s", "m", "l", "xl", "debiased"]
30
+
31
+
32
+ class WinograndeConfig(datasets.BuilderConfig):
33
+
34
+ """BuilderConfig for Discofuse"""
35
+
36
+ def __init__(self, data_size, **kwargs):
37
+ """
38
+
39
+ Args:
40
+ data_size: the format of the training set we want to use (xs, s, m, l, xl, debiased)
41
+ **kwargs: keyword arguments forwarded to super.
42
+ """
43
+ super(WinograndeConfig, self).__init__(version=datasets.Version("1.1.0", ""), **kwargs)
44
+ self.data_size = data_size
45
+
46
+
47
+ class Winogrande(datasets.GeneratorBasedBuilder):
48
+ """TODO(winogrande): Short description of my dataset."""
49
+
50
+ # TODO(winogrande): Set up version.
51
+ VERSION = datasets.Version("1.1.0")
52
+ BUILDER_CONFIGS = [
53
+ WinograndeConfig(name="winogrande_" + data_size, description="AI2 dataset", data_size=data_size)
54
+ for data_size in _FORMATS
55
+ ]
56
+
57
+ def _info(self):
58
+ # TODO(winogrande): Specifies the datasets.DatasetInfo object
59
+ return datasets.DatasetInfo(
60
+ # This is the description that will appear on the datasets page.
61
+ description=_DESCRIPTION,
62
+ # datasets.features.FeatureConnectors
63
+ features=datasets.Features(
64
+ {
65
+ "sentence": datasets.Value("string"),
66
+ "option1": datasets.Value("string"),
67
+ "option2": datasets.Value("string"),
68
+ "answer": datasets.Value("string")
69
+ # These are the features of your dataset like images, labels ...
70
+ }
71
+ ),
72
+ # If there's a common (input, target) tuple from the features,
73
+ # specify them here. They'll be used if as_supervised=True in
74
+ # builder.as_dataset.
75
+ supervised_keys=None,
76
+ # Homepage of the dataset for documentation
77
+ homepage="https://leaderboard.allenai.org/winogrande/submissions/get-started",
78
+ citation=_CITATION,
79
+ )
80
+
81
+ def _split_generators(self, dl_manager):
82
+ """Returns SplitGenerators."""
83
+ # TODO(winogrande): Downloads the data and defines the splits
84
+ # dl_manager is a datasets.download.DownloadManager that can be used to
85
+ # download and extract URLs
86
+ dl_dir = dl_manager.download_and_extract(_URL)
87
+ data_dir = os.path.join(dl_dir, "winogrande_1.1")
88
+ return [
89
+ datasets.SplitGenerator(
90
+ name=datasets.Split.TRAIN,
91
+ # These kwargs will be passed to _generate_examples
92
+ gen_kwargs={
93
+ "filepath": os.path.join(data_dir, f"train_{self.config.data_size}.jsonl"),
94
+ # 'labelpath': os.path.join(data_dir, 'train_{}-labels.lst'.format(self.config.data_size)),
95
+ "split": "train",
96
+ },
97
+ ),
98
+ datasets.SplitGenerator(
99
+ name=datasets.Split.TEST,
100
+ # These kwargs will be passed to _generate_examples
101
+ gen_kwargs={"filepath": os.path.join(data_dir, "test.jsonl"), "split": "test"},
102
+ ),
103
+ datasets.SplitGenerator(
104
+ name=datasets.Split.VALIDATION,
105
+ # These kwargs will be passed to _generate_examples
106
+ gen_kwargs={
107
+ "filepath": os.path.join(data_dir, "dev.jsonl"),
108
+ # 'labelpath': os.path.join(data_dir, 'dev-labels.lst'),
109
+ "split": "dev",
110
+ },
111
+ ),
112
+ ]
113
+
114
+ def _generate_examples(self, filepath, split):
115
+ """Yields examples."""
116
+ # TODO(winogrande): Yields (key, example) tuples from the dataset
117
+ with open(filepath, encoding="utf-8") as f:
118
+ for id_, row in enumerate(f):
119
+ data = json.loads(row)
120
+ if split == "test":
121
+ yield id_, {
122
+ "sentence": data["sentence"],
123
+ "option1": data["option1"],
124
+ "option2": data["option2"],
125
+ "answer": "",
126
+ }
127
+ else:
128
+ yield id_, {
129
+ "sentence": data["sentence"],
130
+ "option1": data["option1"],
131
+ "option2": data["option2"],
132
+ "answer": data["answer"],
133
+ }
134
+
135
+
136
+ # def _generate_test_example(filepath, split, labelpath=None):
137
+ # with open(filepath, encoding="utf-8") as f:
138
+ # for id_, row in enumerate(f):
139
+ # data = json.loads(row)
140
+ # yield id_,{
141
+ # 'sentence': data['sentence'],
142
+ # 'option1': data['option1'],
143
+ # 'option2': data['option2'],
144
+ # 'answer': None
145
+ # }