Hugging Face's logo

Datasets:
laughatwill
/
TIGER-Lab_MMEB-train

Modalities:
Image
Text
Formats:
lance
Languages:
English
Size:
1M - 10M
ArXiv:
Tags:
embedding
lance
multimodal
Libraries:
Datasets
Lance
License:
Dataset card Data Studio Files
xet
Community
laughatwill commited on
Commit
8e2e654
·
verified ·
1 Parent(s): 05ed204

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +146 -362
README.md CHANGED
@@ -9,383 +9,167 @@ tags:
9
  - embedding
10
  - lance
11
  - multimodal
12
- dataset_info:
13
  - config_name: A-OKVQA
14
- features:
15
- - name: qry
16
- dtype: string
17
- - name: qry_image_id
18
- dtype: string
19
- - name: pos_text
20
- dtype: string
21
- - name: pos_image_id
22
- dtype: string
23
- - name: neg_text
24
- dtype: string
25
- - name: neg_image_id
26
- dtype: string
27
- splits:
28
- - name: train
29
- num_examples: 17056
30
- - name: original
31
- num_examples: 17056
32
- - name: diverse_instruction
33
- num_examples: 17056
34
  - config_name: ChartQA
35
- features:
36
- - name: qry
37
- dtype: string
38
- - name: qry_image_id
39
- dtype: string
40
- - name: pos_text
41
- dtype: string
42
- - name: pos_image_id
43
- dtype: string
44
- - name: neg_text
45
- dtype: string
46
- - name: neg_image_id
47
- dtype: string
48
- splits:
49
- - name: train
50
- num_examples: 28299
51
- - name: original
52
- num_examples: 28299
53
- - name: diverse_instruction
54
- num_examples: 28299
55
  - config_name: CIRR
56
- features:
57
- - name: qry
58
- dtype: string
59
- - name: qry_image_id
60
- dtype: string
61
- - name: pos_text
62
- dtype: string
63
- - name: pos_image_id
64
- dtype: string
65
- splits:
66
- - name: train
67
- num_examples: 26116
68
- - name: original
69
- num_examples: 26116
70
- - name: diverse_instruction
71
- num_examples: 26116
72
  - config_name: DocVQA
73
- features:
74
- - name: qry
75
- dtype: string
76
- - name: qry_image_id
77
- dtype: string
78
- - name: pos_text
79
- dtype: string
80
- - name: pos_image_id
81
- dtype: string
82
- - name: neg_text
83
- dtype: string
84
- - name: neg_image_id
85
- dtype: string
86
- splits:
87
- - name: train
88
- num_examples: 39463
89
- - name: original
90
- num_examples: 39463
91
- - name: diverse_instruction
92
- num_examples: 39463
93
  - config_name: HatefulMemes
94
- features:
95
- - name: qry
96
- dtype: string
97
- - name: qry_image_id
98
- dtype: string
99
- - name: pos_text
100
- dtype: string
101
- - name: pos_image_id
102
- dtype: string
103
- - name: neg_text
104
- dtype: string
105
- - name: neg_image_id
106
- dtype: string
107
- splits:
108
- - name: train
109
- num_examples: 8500
110
- - name: original
111
- num_examples: 8500
112
- - name: diverse_instruction
113
- num_examples: 8500
114
  - config_name: ImageNet_1K
115
- features:
116
- - name: qry
117
- dtype: string
118
- - name: qry_image_id
119
- dtype: string
120
- - name: pos_text
121
- dtype: string
122
- - name: pos_image_id
123
- dtype: string
124
- splits:
125
- - name: train
126
- num_examples: 100000
127
- - name: original
128
- num_examples: 100000
129
- - name: diverse_instruction
130
- num_examples: 100000
131
  - config_name: InfographicsVQA
132
- features:
133
- - name: qry
134
- dtype: string
135
- - name: qry_image_id
136
- dtype: string
137
- - name: pos_text
138
- dtype: string
139
- - name: pos_image_id
140
- dtype: string
141
- - name: neg_text
142
- dtype: string
143
- - name: neg_image_id
144
- dtype: string
145
- splits:
146
- - name: train
147
- num_examples: 23946
148
- - name: original
149
- num_examples: 23946
150
- - name: diverse_instruction
151
- num_examples: 23946
152
  - config_name: MSCOCO
153
- features:
154
- - name: qry
155
- dtype: string
156
- - name: qry_image_id
157
- dtype: string
158
- - name: pos_text
159
- dtype: string
160
- - name: pos_image_id
161
- dtype: string
162
- splits:
163
- - name: train
164
- num_examples: 100000
165
- - name: original
166
- num_examples: 100000
167
- - name: diverse_instruction
168
- num_examples: 100000
169
  - config_name: MSCOCO_i2t
170
- features:
171
- - name: qry
172
- dtype: string
173
- - name: qry_image_id
174
- dtype: string
175
- - name: pos_text
176
- dtype: string
177
- - name: pos_image_id
178
- dtype: string
179
- splits:
180
- - name: train
181
- num_examples: 113287
182
- - name: original
183
- num_examples: 113287
184
- - name: diverse_instruction
185
- num_examples: 113287
186
  - config_name: MSCOCO_t2i
187
- features:
188
- - name: qry
189
- dtype: string
190
- - name: qry_image_id
191
- dtype: string
192
- - name: pos_text
193
- dtype: string
194
- - name: pos_image_id
195
- dtype: string
196
- splits:
197
- - name: train
198
- num_examples: 100000
199
- - name: original
200
- num_examples: 100000
201
- - name: diverse_instruction
202
- num_examples: 100000
203
  - config_name: N24News
204
- features:
205
- - name: qry
206
- dtype: string
207
- - name: qry_image_id
208
- dtype: string
209
- - name: pos_text
210
- dtype: string
211
- - name: pos_image_id
212
- dtype: string
213
- - name: neg_text
214
- dtype: string
215
- - name: neg_image_id
216
- dtype: string
217
- splits:
218
- - name: train
219
- num_examples: 48988
220
- - name: original
221
- num_examples: 48988
222
- - name: diverse_instruction
223
- num_examples: 48988
224
  - config_name: NIGHTS
225
- features:
226
- - name: qry
227
- dtype: string
228
- - name: qry_image_id
229
- dtype: string
230
- - name: pos_text
231
- dtype: string
232
- - name: pos_image_id
233
- dtype: string
234
- splits:
235
- - name: train
236
- num_examples: 15941
237
- - name: original
238
- num_examples: 15941
239
- - name: diverse_instruction
240
- num_examples: 15941
241
  - config_name: OK-VQA
242
- features:
243
- - name: qry
244
- dtype: string
245
- - name: qry_image_id
246
- dtype: string
247
- - name: pos_text
248
- dtype: string
249
- - name: pos_image_id
250
- dtype: string
251
- - name: neg_text
252
- dtype: string
253
- - name: neg_image_id
254
- dtype: string
255
- splits:
256
- - name: train
257
- num_examples: 9009
258
- - name: original
259
- num_examples: 9009
260
- - name: diverse_instruction
261
- num_examples: 9009
262
  - config_name: SUN397
263
- features:
264
- - name: qry
265
- dtype: string
266
- - name: qry_image_id
267
- dtype: string
268
- - name: pos_text
269
- dtype: string
270
- - name: pos_image_id
271
- dtype: string
272
- splits:
273
- - name: train
274
- num_examples: 19850
275
- - name: original
276
- num_examples: 19850
277
- - name: diverse_instruction
278
- num_examples: 19850
279
  - config_name: VOC2007
280
- features:
281
- - name: qry
282
- dtype: string
283
- - name: qry_image_id
284
- dtype: string
285
- - name: pos_text
286
- dtype: string
287
- - name: pos_image_id
288
- dtype: string
289
- splits:
290
- - name: train
291
- num_examples: 7844
292
- - name: original
293
- num_examples: 7844
294
- - name: diverse_instruction
295
- num_examples: 7844
296
  - config_name: VisDial
297
- features:
298
- - name: qry
299
- dtype: string
300
- - name: qry_image_id
301
- dtype: string
302
- - name: pos_text
303
- dtype: string
304
- - name: pos_image_id
305
- dtype: string
306
- - name: neg_text
307
- dtype: string
308
- - name: neg_image_id
309
- dtype: string
310
- splits:
311
- - name: train
312
- num_examples: 123287
313
- - name: original
314
- num_examples: 123287
315
- - name: diverse_instruction
316
- num_examples: 123287
317
  - config_name: Visual7W
318
- features:
319
- - name: qry
320
- dtype: string
321
- - name: qry_image_id
322
- dtype: string
323
- - name: pos_text
324
- dtype: string
325
- - name: pos_image_id
326
- dtype: string
327
- - name: neg_text
328
- dtype: string
329
- - name: neg_image_id
330
- dtype: string
331
- splits:
332
- - name: train
333
- num_examples: 69817
334
- - name: original
335
- num_examples: 69817
336
- - name: diverse_instruction
337
- num_examples: 69817
338
  - config_name: VisualNews_i2t
339
- features:
340
- - name: qry
341
- dtype: string
342
- - name: qry_image_id
343
- dtype: string
344
- - name: pos_text
345
- dtype: string
346
- - name: pos_image_id
347
- dtype: string
348
- splits:
349
- - name: train
350
- num_examples: 100000
351
- - name: original
352
- num_examples: 100000
353
- - name: diverse_instruction
354
- num_examples: 100000
355
  - config_name: VisualNews_t2i
356
- features:
357
- - name: qry
358
- dtype: string
359
- - name: qry_image_id
360
- dtype: string
361
- - name: pos_text
362
- dtype: string
363
- - name: pos_image_id
364
- dtype: string
365
- splits:
366
- - name: train
367
- num_examples: 99903
368
- - name: original
369
- num_examples: 99903
370
- - name: diverse_instruction
371
- num_examples: 99903
372
  - config_name: WebQA
373
- features:
374
- - name: qry
375
- dtype: string
376
- - name: qry_image_id
377
- dtype: string
378
- - name: pos_text
379
- dtype: string
380
- - name: pos_image_id
381
- dtype: string
382
- splits:
383
- - name: train
384
- num_examples: 17166
385
- - name: original
386
- num_examples: 17166
387
- - name: diverse_instruction
388
- num_examples: 17166
389
  ---
390
 
391
  # MMEB Training Dataset (Lance Format)
@@ -400,14 +184,14 @@ The original dataset is used for training VLM2Vec models in the paper [VLM2Vec:
400
  TIGER-Lab_MMEB-train/
401
  └── data/
402
  ├── A-OKVQA/
403
- │ ├── train.lance/ # Training metadata
404
- │ ├── original.lance/ # Original instructions
405
- │ └── diverse.lance/ # Diverse instructions
406
  ├── MSCOCO/
407
  │ └── ...
408
  └── images/
409
- ├── A-OKVQA.lance/ # Images (binary)
410
- ├── MSCOCO.lance/
411
  └── ...
412
  ```
413
 
 
9
  - embedding
10
  - lance
11
  - multimodal
12
+ configs:
13
  - config_name: A-OKVQA
14
+ data_files:
15
+ - split: train
16
+ path: data/A-OKVQA/train.lance
17
+ - split: original
18
+ path: data/A-OKVQA/original.lance
19
+ - split: diverse_instruction
20
+ path: data/A-OKVQA/diverse.lance
 
 
 
 
 
 
 
 
 
 
 
 
 
21
  - config_name: ChartQA
22
+ data_files:
23
+ - split: train
24
+ path: data/ChartQA/train.lance
25
+ - split: original
26
+ path: data/ChartQA/original.lance
27
+ - split: diverse_instruction
28
+ path: data/ChartQA/diverse.lance
 
 
 
 
 
 
 
 
 
 
 
 
 
29
  - config_name: CIRR
30
+ data_files:
31
+ - split: train
32
+ path: data/CIRR/train.lance
33
+ - split: original
34
+ path: data/CIRR/original.lance
35
+ - split: diverse_instruction
36
+ path: data/CIRR/diverse.lance
 
 
 
 
 
 
 
 
 
37
  - config_name: DocVQA
38
+ data_files:
39
+ - split: train
40
+ path: data/DocVQA/train.lance
41
+ - split: original
42
+ path: data/DocVQA/original.lance
43
+ - split: diverse_instruction
44
+ path: data/DocVQA/diverse.lance
 
 
 
 
 
 
 
 
 
 
 
 
 
45
  - config_name: HatefulMemes
46
+ data_files:
47
+ - split: train
48
+ path: data/HatefulMemes/train.lance
49
+ - split: original
50
+ path: data/HatefulMemes/original.lance
51
+ - split: diverse_instruction
52
+ path: data/HatefulMemes/diverse.lance
 
 
 
 
 
 
 
 
 
 
 
 
 
53
  - config_name: ImageNet_1K
54
+ data_files:
55
+ - split: train
56
+ path: data/ImageNet_1K/train.lance
57
+ - split: original
58
+ path: data/ImageNet_1K/original.lance
59
+ - split: diverse_instruction
60
+ path: data/ImageNet_1K/diverse.lance
 
 
 
 
 
 
 
 
 
61
  - config_name: InfographicsVQA
62
+ data_files:
63
+ - split: train
64
+ path: data/InfographicsVQA/train.lance
65
+ - split: original
66
+ path: data/InfographicsVQA/original.lance
67
+ - split: diverse_instruction
68
+ path: data/InfographicsVQA/diverse.lance
 
 
 
 
 
 
 
 
 
 
 
 
 
69
  - config_name: MSCOCO
70
+ data_files:
71
+ - split: train
72
+ path: data/MSCOCO/train.lance
73
+ - split: original
74
+ path: data/MSCOCO/original.lance
75
+ - split: diverse_instruction
76
+ path: data/MSCOCO/diverse.lance
 
 
 
 
 
 
 
 
 
77
  - config_name: MSCOCO_i2t
78
+ data_files:
79
+ - split: train
80
+ path: data/MSCOCO_i2t/train.lance
81
+ - split: original
82
+ path: data/MSCOCO_i2t/original.lance
83
+ - split: diverse_instruction
84
+ path: data/MSCOCO_i2t/diverse.lance
 
 
 
 
 
 
 
 
 
85
  - config_name: MSCOCO_t2i
86
+ data_files:
87
+ - split: train
88
+ path: data/MSCOCO_t2i/train.lance
89
+ - split: original
90
+ path: data/MSCOCO_t2i/original.lance
91
+ - split: diverse_instruction
92
+ path: data/MSCOCO_t2i/diverse.lance
 
 
 
 
 
 
 
 
 
93
  - config_name: N24News
94
+ data_files:
95
+ - split: train
96
+ path: data/N24News/train.lance
97
+ - split: original
98
+ path: data/N24News/original.lance
99
+ - split: diverse_instruction
100
+ path: data/N24News/diverse.lance
 
 
 
 
 
 
 
 
 
 
 
 
 
101
  - config_name: NIGHTS
102
+ data_files:
103
+ - split: train
104
+ path: data/NIGHTS/train.lance
105
+ - split: original
106
+ path: data/NIGHTS/original.lance
107
+ - split: diverse_instruction
108
+ path: data/NIGHTS/diverse.lance
 
 
 
 
 
 
 
 
 
109
  - config_name: OK-VQA
110
+ data_files:
111
+ - split: train
112
+ path: data/OK-VQA/train.lance
113
+ - split: original
114
+ path: data/OK-VQA/original.lance
115
+ - split: diverse_instruction
116
+ path: data/OK-VQA/diverse.lance
 
 
 
 
 
 
 
 
 
 
 
 
 
117
  - config_name: SUN397
118
+ data_files:
119
+ - split: train
120
+ path: data/SUN397/train.lance
121
+ - split: original
122
+ path: data/SUN397/original.lance
123
+ - split: diverse_instruction
124
+ path: data/SUN397/diverse.lance
 
 
 
 
 
 
 
 
 
125
  - config_name: VOC2007
126
+ data_files:
127
+ - split: train
128
+ path: data/VOC2007/train.lance
129
+ - split: original
130
+ path: data/VOC2007/original.lance
131
+ - split: diverse_instruction
132
+ path: data/VOC2007/diverse.lance
 
 
 
 
 
 
 
 
 
133
  - config_name: VisDial
134
+ data_files:
135
+ - split: train
136
+ path: data/VisDial/train.lance
137
+ - split: original
138
+ path: data/VisDial/original.lance
139
+ - split: diverse_instruction
140
+ path: data/VisDial/diverse.lance
 
 
 
 
 
 
 
 
 
 
 
 
 
141
  - config_name: Visual7W
142
+ data_files:
143
+ - split: train
144
+ path: data/Visual7W/train.lance
145
+ - split: original
146
+ path: data/Visual7W/original.lance
147
+ - split: diverse_instruction
148
+ path: data/Visual7W/diverse.lance
 
 
 
 
 
 
 
 
 
 
 
 
 
149
  - config_name: VisualNews_i2t
150
+ data_files:
151
+ - split: train
152
+ path: data/VisualNews_i2t/train.lance
153
+ - split: original
154
+ path: data/VisualNews_i2t/original.lance
155
+ - split: diverse_instruction
156
+ path: data/VisualNews_i2t/diverse.lance
 
 
 
 
 
 
 
 
 
157
  - config_name: VisualNews_t2i
158
+ data_files:
159
+ - split: train
160
+ path: data/VisualNews_t2i/train.lance
161
+ - split: original
162
+ path: data/VisualNews_t2i/original.lance
163
+ - split: diverse_instruction
164
+ path: data/VisualNews_t2i/diverse.lance
 
 
 
 
 
 
 
 
 
165
  - config_name: WebQA
166
+ data_files:
167
+ - split: train
168
+ path: data/WebQA/train.lance
169
+ - split: original
170
+ path: data/WebQA/original.lance
171
+ - split: diverse_instruction
172
+ path: data/WebQA/diverse.lance
 
 
 
 
 
 
 
 
 
173
  ---
174
 
175
  # MMEB Training Dataset (Lance Format)
 
184
  TIGER-Lab_MMEB-train/
185
  └── data/
186
  ├── A-OKVQA/
187
+ │ ├── train.lance
188
+ │ ├── original.lance
189
+ │ └── diverse.lance
190
  ├── MSCOCO/
191
  │ └── ...
192
  └── images/
193
+ ├── A-OKVQA.lance
194
+ ├── MSCOCO.lance
195
  └── ...
196
  ```
197