AlexZigma commited on
Commit
f4c9911
1 Parent(s): 7b2e49a

Upload model

Browse files
Files changed (3) hide show
  1. config.json +974 -0
  2. generation_config.json +5 -0
  3. pytorch_model.bin +3 -0
config.json ADDED
@@ -0,0 +1,974 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_commit_hash": null,
3
+ "_name_or_path": "/content/content/timesformer-bert-video-captioning/checkpoint-1700",
4
+ "architectures": [
5
+ "VisionEncoderDecoderModel"
6
+ ],
7
+ "decoder": {
8
+ "_name_or_path": "bert-base-uncased",
9
+ "add_cross_attention": true,
10
+ "architectures": [
11
+ "BertForMaskedLM"
12
+ ],
13
+ "attention_probs_dropout_prob": 0.1,
14
+ "bad_words_ids": null,
15
+ "begin_suppress_tokens": null,
16
+ "bos_token_id": null,
17
+ "chunk_size_feed_forward": 0,
18
+ "classifier_dropout": null,
19
+ "cross_attention_hidden_size": null,
20
+ "decoder_start_token_id": null,
21
+ "diversity_penalty": 0.0,
22
+ "do_sample": false,
23
+ "early_stopping": false,
24
+ "encoder_no_repeat_ngram_size": 0,
25
+ "eos_token_id": null,
26
+ "exponential_decay_length_penalty": null,
27
+ "finetuning_task": null,
28
+ "forced_bos_token_id": null,
29
+ "forced_eos_token_id": null,
30
+ "gradient_checkpointing": false,
31
+ "hidden_act": "gelu",
32
+ "hidden_dropout_prob": 0.1,
33
+ "hidden_size": 768,
34
+ "id2label": {
35
+ "0": "LABEL_0",
36
+ "1": "LABEL_1"
37
+ },
38
+ "initializer_range": 0.02,
39
+ "intermediate_size": 3072,
40
+ "is_decoder": true,
41
+ "is_encoder_decoder": false,
42
+ "label2id": {
43
+ "LABEL_0": 0,
44
+ "LABEL_1": 1
45
+ },
46
+ "layer_norm_eps": 1e-12,
47
+ "length_penalty": 1.0,
48
+ "max_length": 20,
49
+ "max_position_embeddings": 512,
50
+ "min_length": 0,
51
+ "model_type": "bert",
52
+ "no_repeat_ngram_size": 0,
53
+ "num_attention_heads": 12,
54
+ "num_beam_groups": 1,
55
+ "num_beams": 1,
56
+ "num_hidden_layers": 12,
57
+ "num_return_sequences": 1,
58
+ "output_attentions": false,
59
+ "output_hidden_states": false,
60
+ "output_scores": false,
61
+ "pad_token_id": 0,
62
+ "position_embedding_type": "absolute",
63
+ "prefix": null,
64
+ "problem_type": null,
65
+ "pruned_heads": {},
66
+ "remove_invalid_values": false,
67
+ "repetition_penalty": 1.0,
68
+ "return_dict": true,
69
+ "return_dict_in_generate": false,
70
+ "sep_token_id": null,
71
+ "suppress_tokens": null,
72
+ "task_specific_params": null,
73
+ "temperature": 1.0,
74
+ "tf_legacy_loss": false,
75
+ "tie_encoder_decoder": false,
76
+ "tie_word_embeddings": true,
77
+ "tokenizer_class": null,
78
+ "top_k": 50,
79
+ "top_p": 1.0,
80
+ "torch_dtype": null,
81
+ "torchscript": false,
82
+ "transformers_version": "4.30.2",
83
+ "type_vocab_size": 2,
84
+ "typical_p": 1.0,
85
+ "use_bfloat16": false,
86
+ "use_cache": true,
87
+ "vocab_size": 30522
88
+ },
89
+ "decoder_start_token_id": 101,
90
+ "encoder": {
91
+ "_name_or_path": "facebook/timesformer-base-finetuned-k400",
92
+ "add_cross_attention": false,
93
+ "architectures": [
94
+ "TimesformerForVideoClassification"
95
+ ],
96
+ "attention_probs_dropout_prob": 0.0,
97
+ "attention_type": "divided_space_time",
98
+ "bad_words_ids": null,
99
+ "begin_suppress_tokens": null,
100
+ "bos_token_id": null,
101
+ "chunk_size_feed_forward": 0,
102
+ "cross_attention_hidden_size": null,
103
+ "decoder_start_token_id": null,
104
+ "diversity_penalty": 0.0,
105
+ "do_sample": false,
106
+ "drop_path_rate": 0,
107
+ "early_stopping": false,
108
+ "encoder_no_repeat_ngram_size": 0,
109
+ "eos_token_id": null,
110
+ "exponential_decay_length_penalty": null,
111
+ "finetuning_task": null,
112
+ "forced_bos_token_id": null,
113
+ "forced_eos_token_id": null,
114
+ "hidden_act": "gelu",
115
+ "hidden_dropout_prob": 0.0,
116
+ "hidden_size": 768,
117
+ "id2label": {
118
+ "0": "abseiling",
119
+ "1": "air drumming",
120
+ "2": "answering questions",
121
+ "3": "applauding",
122
+ "4": "applying cream",
123
+ "5": "archery",
124
+ "6": "arm wrestling",
125
+ "7": "arranging flowers",
126
+ "8": "assembling computer",
127
+ "9": "auctioning",
128
+ "10": "baby waking up",
129
+ "11": "baking cookies",
130
+ "12": "balloon blowing",
131
+ "13": "bandaging",
132
+ "14": "barbequing",
133
+ "15": "bartending",
134
+ "16": "beatboxing",
135
+ "17": "bee keeping",
136
+ "18": "belly dancing",
137
+ "19": "bench pressing",
138
+ "20": "bending back",
139
+ "21": "bending metal",
140
+ "22": "biking through snow",
141
+ "23": "blasting sand",
142
+ "24": "blowing glass",
143
+ "25": "blowing leaves",
144
+ "26": "blowing nose",
145
+ "27": "blowing out candles",
146
+ "28": "bobsledding",
147
+ "29": "bookbinding",
148
+ "30": "bouncing on trampoline",
149
+ "31": "bowling",
150
+ "32": "braiding hair",
151
+ "33": "breading or breadcrumbing",
152
+ "34": "breakdancing",
153
+ "35": "brush painting",
154
+ "36": "brushing hair",
155
+ "37": "brushing teeth",
156
+ "38": "building cabinet",
157
+ "39": "building shed",
158
+ "40": "bungee jumping",
159
+ "41": "busking",
160
+ "42": "canoeing or kayaking",
161
+ "43": "capoeira",
162
+ "44": "carrying baby",
163
+ "45": "cartwheeling",
164
+ "46": "carving pumpkin",
165
+ "47": "catching fish",
166
+ "48": "catching or throwing baseball",
167
+ "49": "catching or throwing frisbee",
168
+ "50": "catching or throwing softball",
169
+ "51": "celebrating",
170
+ "52": "changing oil",
171
+ "53": "changing wheel",
172
+ "54": "checking tires",
173
+ "55": "cheerleading",
174
+ "56": "chopping wood",
175
+ "57": "clapping",
176
+ "58": "clay pottery making",
177
+ "59": "clean and jerk",
178
+ "60": "cleaning floor",
179
+ "61": "cleaning gutters",
180
+ "62": "cleaning pool",
181
+ "63": "cleaning shoes",
182
+ "64": "cleaning toilet",
183
+ "65": "cleaning windows",
184
+ "66": "climbing a rope",
185
+ "67": "climbing ladder",
186
+ "68": "climbing tree",
187
+ "69": "contact juggling",
188
+ "70": "cooking chicken",
189
+ "71": "cooking egg",
190
+ "72": "cooking on campfire",
191
+ "73": "cooking sausages",
192
+ "74": "counting money",
193
+ "75": "country line dancing",
194
+ "76": "cracking neck",
195
+ "77": "crawling baby",
196
+ "78": "crossing river",
197
+ "79": "crying",
198
+ "80": "curling hair",
199
+ "81": "cutting nails",
200
+ "82": "cutting pineapple",
201
+ "83": "cutting watermelon",
202
+ "84": "dancing ballet",
203
+ "85": "dancing charleston",
204
+ "86": "dancing gangnam style",
205
+ "87": "dancing macarena",
206
+ "88": "deadlifting",
207
+ "89": "decorating the christmas tree",
208
+ "90": "digging",
209
+ "91": "dining",
210
+ "92": "disc golfing",
211
+ "93": "diving cliff",
212
+ "94": "dodgeball",
213
+ "95": "doing aerobics",
214
+ "96": "doing laundry",
215
+ "97": "doing nails",
216
+ "98": "drawing",
217
+ "99": "dribbling basketball",
218
+ "100": "drinking",
219
+ "101": "drinking beer",
220
+ "102": "drinking shots",
221
+ "103": "driving car",
222
+ "104": "driving tractor",
223
+ "105": "drop kicking",
224
+ "106": "drumming fingers",
225
+ "107": "dunking basketball",
226
+ "108": "dying hair",
227
+ "109": "eating burger",
228
+ "110": "eating cake",
229
+ "111": "eating carrots",
230
+ "112": "eating chips",
231
+ "113": "eating doughnuts",
232
+ "114": "eating hotdog",
233
+ "115": "eating ice cream",
234
+ "116": "eating spaghetti",
235
+ "117": "eating watermelon",
236
+ "118": "egg hunting",
237
+ "119": "exercising arm",
238
+ "120": "exercising with an exercise ball",
239
+ "121": "extinguishing fire",
240
+ "122": "faceplanting",
241
+ "123": "feeding birds",
242
+ "124": "feeding fish",
243
+ "125": "feeding goats",
244
+ "126": "filling eyebrows",
245
+ "127": "finger snapping",
246
+ "128": "fixing hair",
247
+ "129": "flipping pancake",
248
+ "130": "flying kite",
249
+ "131": "folding clothes",
250
+ "132": "folding napkins",
251
+ "133": "folding paper",
252
+ "134": "front raises",
253
+ "135": "frying vegetables",
254
+ "136": "garbage collecting",
255
+ "137": "gargling",
256
+ "138": "getting a haircut",
257
+ "139": "getting a tattoo",
258
+ "140": "giving or receiving award",
259
+ "141": "golf chipping",
260
+ "142": "golf driving",
261
+ "143": "golf putting",
262
+ "144": "grinding meat",
263
+ "145": "grooming dog",
264
+ "146": "grooming horse",
265
+ "147": "gymnastics tumbling",
266
+ "148": "hammer throw",
267
+ "149": "headbanging",
268
+ "150": "headbutting",
269
+ "151": "high jump",
270
+ "152": "high kick",
271
+ "153": "hitting baseball",
272
+ "154": "hockey stop",
273
+ "155": "holding snake",
274
+ "156": "hopscotch",
275
+ "157": "hoverboarding",
276
+ "158": "hugging",
277
+ "159": "hula hooping",
278
+ "160": "hurdling",
279
+ "161": "hurling (sport)",
280
+ "162": "ice climbing",
281
+ "163": "ice fishing",
282
+ "164": "ice skating",
283
+ "165": "ironing",
284
+ "166": "javelin throw",
285
+ "167": "jetskiing",
286
+ "168": "jogging",
287
+ "169": "juggling balls",
288
+ "170": "juggling fire",
289
+ "171": "juggling soccer ball",
290
+ "172": "jumping into pool",
291
+ "173": "jumpstyle dancing",
292
+ "174": "kicking field goal",
293
+ "175": "kicking soccer ball",
294
+ "176": "kissing",
295
+ "177": "kitesurfing",
296
+ "178": "knitting",
297
+ "179": "krumping",
298
+ "180": "laughing",
299
+ "181": "laying bricks",
300
+ "182": "long jump",
301
+ "183": "lunge",
302
+ "184": "making a cake",
303
+ "185": "making a sandwich",
304
+ "186": "making bed",
305
+ "187": "making jewelry",
306
+ "188": "making pizza",
307
+ "189": "making snowman",
308
+ "190": "making sushi",
309
+ "191": "making tea",
310
+ "192": "marching",
311
+ "193": "massaging back",
312
+ "194": "massaging feet",
313
+ "195": "massaging legs",
314
+ "196": "massaging person's head",
315
+ "197": "milking cow",
316
+ "198": "mopping floor",
317
+ "199": "motorcycling",
318
+ "200": "moving furniture",
319
+ "201": "mowing lawn",
320
+ "202": "news anchoring",
321
+ "203": "opening bottle",
322
+ "204": "opening present",
323
+ "205": "paragliding",
324
+ "206": "parasailing",
325
+ "207": "parkour",
326
+ "208": "passing American football (in game)",
327
+ "209": "passing American football (not in game)",
328
+ "210": "peeling apples",
329
+ "211": "peeling potatoes",
330
+ "212": "petting animal (not cat)",
331
+ "213": "petting cat",
332
+ "214": "picking fruit",
333
+ "215": "planting trees",
334
+ "216": "plastering",
335
+ "217": "playing accordion",
336
+ "218": "playing badminton",
337
+ "219": "playing bagpipes",
338
+ "220": "playing basketball",
339
+ "221": "playing bass guitar",
340
+ "222": "playing cards",
341
+ "223": "playing cello",
342
+ "224": "playing chess",
343
+ "225": "playing clarinet",
344
+ "226": "playing controller",
345
+ "227": "playing cricket",
346
+ "228": "playing cymbals",
347
+ "229": "playing didgeridoo",
348
+ "230": "playing drums",
349
+ "231": "playing flute",
350
+ "232": "playing guitar",
351
+ "233": "playing harmonica",
352
+ "234": "playing harp",
353
+ "235": "playing ice hockey",
354
+ "236": "playing keyboard",
355
+ "237": "playing kickball",
356
+ "238": "playing monopoly",
357
+ "239": "playing organ",
358
+ "240": "playing paintball",
359
+ "241": "playing piano",
360
+ "242": "playing poker",
361
+ "243": "playing recorder",
362
+ "244": "playing saxophone",
363
+ "245": "playing squash or racquetball",
364
+ "246": "playing tennis",
365
+ "247": "playing trombone",
366
+ "248": "playing trumpet",
367
+ "249": "playing ukulele",
368
+ "250": "playing violin",
369
+ "251": "playing volleyball",
370
+ "252": "playing xylophone",
371
+ "253": "pole vault",
372
+ "254": "presenting weather forecast",
373
+ "255": "pull ups",
374
+ "256": "pumping fist",
375
+ "257": "pumping gas",
376
+ "258": "punching bag",
377
+ "259": "punching person (boxing)",
378
+ "260": "push up",
379
+ "261": "pushing car",
380
+ "262": "pushing cart",
381
+ "263": "pushing wheelchair",
382
+ "264": "reading book",
383
+ "265": "reading newspaper",
384
+ "266": "recording music",
385
+ "267": "riding a bike",
386
+ "268": "riding camel",
387
+ "269": "riding elephant",
388
+ "270": "riding mechanical bull",
389
+ "271": "riding mountain bike",
390
+ "272": "riding mule",
391
+ "273": "riding or walking with horse",
392
+ "274": "riding scooter",
393
+ "275": "riding unicycle",
394
+ "276": "ripping paper",
395
+ "277": "robot dancing",
396
+ "278": "rock climbing",
397
+ "279": "rock scissors paper",
398
+ "280": "roller skating",
399
+ "281": "running on treadmill",
400
+ "282": "sailing",
401
+ "283": "salsa dancing",
402
+ "284": "sanding floor",
403
+ "285": "scrambling eggs",
404
+ "286": "scuba diving",
405
+ "287": "setting table",
406
+ "288": "shaking hands",
407
+ "289": "shaking head",
408
+ "290": "sharpening knives",
409
+ "291": "sharpening pencil",
410
+ "292": "shaving head",
411
+ "293": "shaving legs",
412
+ "294": "shearing sheep",
413
+ "295": "shining shoes",
414
+ "296": "shooting basketball",
415
+ "297": "shooting goal (soccer)",
416
+ "298": "shot put",
417
+ "299": "shoveling snow",
418
+ "300": "shredding paper",
419
+ "301": "shuffling cards",
420
+ "302": "side kick",
421
+ "303": "sign language interpreting",
422
+ "304": "singing",
423
+ "305": "situp",
424
+ "306": "skateboarding",
425
+ "307": "ski jumping",
426
+ "308": "skiing (not slalom or crosscountry)",
427
+ "309": "skiing crosscountry",
428
+ "310": "skiing slalom",
429
+ "311": "skipping rope",
430
+ "312": "skydiving",
431
+ "313": "slacklining",
432
+ "314": "slapping",
433
+ "315": "sled dog racing",
434
+ "316": "smoking",
435
+ "317": "smoking hookah",
436
+ "318": "snatch weight lifting",
437
+ "319": "sneezing",
438
+ "320": "sniffing",
439
+ "321": "snorkeling",
440
+ "322": "snowboarding",
441
+ "323": "snowkiting",
442
+ "324": "snowmobiling",
443
+ "325": "somersaulting",
444
+ "326": "spinning poi",
445
+ "327": "spray painting",
446
+ "328": "spraying",
447
+ "329": "springboard diving",
448
+ "330": "squat",
449
+ "331": "sticking tongue out",
450
+ "332": "stomping grapes",
451
+ "333": "stretching arm",
452
+ "334": "stretching leg",
453
+ "335": "strumming guitar",
454
+ "336": "surfing crowd",
455
+ "337": "surfing water",
456
+ "338": "sweeping floor",
457
+ "339": "swimming backstroke",
458
+ "340": "swimming breast stroke",
459
+ "341": "swimming butterfly stroke",
460
+ "342": "swing dancing",
461
+ "343": "swinging legs",
462
+ "344": "swinging on something",
463
+ "345": "sword fighting",
464
+ "346": "tai chi",
465
+ "347": "taking a shower",
466
+ "348": "tango dancing",
467
+ "349": "tap dancing",
468
+ "350": "tapping guitar",
469
+ "351": "tapping pen",
470
+ "352": "tasting beer",
471
+ "353": "tasting food",
472
+ "354": "testifying",
473
+ "355": "texting",
474
+ "356": "throwing axe",
475
+ "357": "throwing ball",
476
+ "358": "throwing discus",
477
+ "359": "tickling",
478
+ "360": "tobogganing",
479
+ "361": "tossing coin",
480
+ "362": "tossing salad",
481
+ "363": "training dog",
482
+ "364": "trapezing",
483
+ "365": "trimming or shaving beard",
484
+ "366": "trimming trees",
485
+ "367": "triple jump",
486
+ "368": "tying bow tie",
487
+ "369": "tying knot (not on a tie)",
488
+ "370": "tying tie",
489
+ "371": "unboxing",
490
+ "372": "unloading truck",
491
+ "373": "using computer",
492
+ "374": "using remote controller (not gaming)",
493
+ "375": "using segway",
494
+ "376": "vault",
495
+ "377": "waiting in line",
496
+ "378": "walking the dog",
497
+ "379": "washing dishes",
498
+ "380": "washing feet",
499
+ "381": "washing hair",
500
+ "382": "washing hands",
501
+ "383": "water skiing",
502
+ "384": "water sliding",
503
+ "385": "watering plants",
504
+ "386": "waxing back",
505
+ "387": "waxing chest",
506
+ "388": "waxing eyebrows",
507
+ "389": "waxing legs",
508
+ "390": "weaving basket",
509
+ "391": "welding",
510
+ "392": "whistling",
511
+ "393": "windsurfing",
512
+ "394": "wrapping present",
513
+ "395": "wrestling",
514
+ "396": "writing",
515
+ "397": "yawning",
516
+ "398": "yoga",
517
+ "399": "zumba"
518
+ },
519
+ "image_size": 224,
520
+ "initializer_range": 0.02,
521
+ "intermediate_size": 3072,
522
+ "is_decoder": false,
523
+ "is_encoder_decoder": false,
524
+ "label2id": {
525
+ "abseiling": 0,
526
+ "air drumming": 1,
527
+ "answering questions": 2,
528
+ "applauding": 3,
529
+ "applying cream": 4,
530
+ "archery": 5,
531
+ "arm wrestling": 6,
532
+ "arranging flowers": 7,
533
+ "assembling computer": 8,
534
+ "auctioning": 9,
535
+ "baby waking up": 10,
536
+ "baking cookies": 11,
537
+ "balloon blowing": 12,
538
+ "bandaging": 13,
539
+ "barbequing": 14,
540
+ "bartending": 15,
541
+ "beatboxing": 16,
542
+ "bee keeping": 17,
543
+ "belly dancing": 18,
544
+ "bench pressing": 19,
545
+ "bending back": 20,
546
+ "bending metal": 21,
547
+ "biking through snow": 22,
548
+ "blasting sand": 23,
549
+ "blowing glass": 24,
550
+ "blowing leaves": 25,
551
+ "blowing nose": 26,
552
+ "blowing out candles": 27,
553
+ "bobsledding": 28,
554
+ "bookbinding": 29,
555
+ "bouncing on trampoline": 30,
556
+ "bowling": 31,
557
+ "braiding hair": 32,
558
+ "breading or breadcrumbing": 33,
559
+ "breakdancing": 34,
560
+ "brush painting": 35,
561
+ "brushing hair": 36,
562
+ "brushing teeth": 37,
563
+ "building cabinet": 38,
564
+ "building shed": 39,
565
+ "bungee jumping": 40,
566
+ "busking": 41,
567
+ "canoeing or kayaking": 42,
568
+ "capoeira": 43,
569
+ "carrying baby": 44,
570
+ "cartwheeling": 45,
571
+ "carving pumpkin": 46,
572
+ "catching fish": 47,
573
+ "catching or throwing baseball": 48,
574
+ "catching or throwing frisbee": 49,
575
+ "catching or throwing softball": 50,
576
+ "celebrating": 51,
577
+ "changing oil": 52,
578
+ "changing wheel": 53,
579
+ "checking tires": 54,
580
+ "cheerleading": 55,
581
+ "chopping wood": 56,
582
+ "clapping": 57,
583
+ "clay pottery making": 58,
584
+ "clean and jerk": 59,
585
+ "cleaning floor": 60,
586
+ "cleaning gutters": 61,
587
+ "cleaning pool": 62,
588
+ "cleaning shoes": 63,
589
+ "cleaning toilet": 64,
590
+ "cleaning windows": 65,
591
+ "climbing a rope": 66,
592
+ "climbing ladder": 67,
593
+ "climbing tree": 68,
594
+ "contact juggling": 69,
595
+ "cooking chicken": 70,
596
+ "cooking egg": 71,
597
+ "cooking on campfire": 72,
598
+ "cooking sausages": 73,
599
+ "counting money": 74,
600
+ "country line dancing": 75,
601
+ "cracking neck": 76,
602
+ "crawling baby": 77,
603
+ "crossing river": 78,
604
+ "crying": 79,
605
+ "curling hair": 80,
606
+ "cutting nails": 81,
607
+ "cutting pineapple": 82,
608
+ "cutting watermelon": 83,
609
+ "dancing ballet": 84,
610
+ "dancing charleston": 85,
611
+ "dancing gangnam style": 86,
612
+ "dancing macarena": 87,
613
+ "deadlifting": 88,
614
+ "decorating the christmas tree": 89,
615
+ "digging": 90,
616
+ "dining": 91,
617
+ "disc golfing": 92,
618
+ "diving cliff": 93,
619
+ "dodgeball": 94,
620
+ "doing aerobics": 95,
621
+ "doing laundry": 96,
622
+ "doing nails": 97,
623
+ "drawing": 98,
624
+ "dribbling basketball": 99,
625
+ "drinking": 100,
626
+ "drinking beer": 101,
627
+ "drinking shots": 102,
628
+ "driving car": 103,
629
+ "driving tractor": 104,
630
+ "drop kicking": 105,
631
+ "drumming fingers": 106,
632
+ "dunking basketball": 107,
633
+ "dying hair": 108,
634
+ "eating burger": 109,
635
+ "eating cake": 110,
636
+ "eating carrots": 111,
637
+ "eating chips": 112,
638
+ "eating doughnuts": 113,
639
+ "eating hotdog": 114,
640
+ "eating ice cream": 115,
641
+ "eating spaghetti": 116,
642
+ "eating watermelon": 117,
643
+ "egg hunting": 118,
644
+ "exercising arm": 119,
645
+ "exercising with an exercise ball": 120,
646
+ "extinguishing fire": 121,
647
+ "faceplanting": 122,
648
+ "feeding birds": 123,
649
+ "feeding fish": 124,
650
+ "feeding goats": 125,
651
+ "filling eyebrows": 126,
652
+ "finger snapping": 127,
653
+ "fixing hair": 128,
654
+ "flipping pancake": 129,
655
+ "flying kite": 130,
656
+ "folding clothes": 131,
657
+ "folding napkins": 132,
658
+ "folding paper": 133,
659
+ "front raises": 134,
660
+ "frying vegetables": 135,
661
+ "garbage collecting": 136,
662
+ "gargling": 137,
663
+ "getting a haircut": 138,
664
+ "getting a tattoo": 139,
665
+ "giving or receiving award": 140,
666
+ "golf chipping": 141,
667
+ "golf driving": 142,
668
+ "golf putting": 143,
669
+ "grinding meat": 144,
670
+ "grooming dog": 145,
671
+ "grooming horse": 146,
672
+ "gymnastics tumbling": 147,
673
+ "hammer throw": 148,
674
+ "headbanging": 149,
675
+ "headbutting": 150,
676
+ "high jump": 151,
677
+ "high kick": 152,
678
+ "hitting baseball": 153,
679
+ "hockey stop": 154,
680
+ "holding snake": 155,
681
+ "hopscotch": 156,
682
+ "hoverboarding": 157,
683
+ "hugging": 158,
684
+ "hula hooping": 159,
685
+ "hurdling": 160,
686
+ "hurling (sport)": 161,
687
+ "ice climbing": 162,
688
+ "ice fishing": 163,
689
+ "ice skating": 164,
690
+ "ironing": 165,
691
+ "javelin throw": 166,
692
+ "jetskiing": 167,
693
+ "jogging": 168,
694
+ "juggling balls": 169,
695
+ "juggling fire": 170,
696
+ "juggling soccer ball": 171,
697
+ "jumping into pool": 172,
698
+ "jumpstyle dancing": 173,
699
+ "kicking field goal": 174,
700
+ "kicking soccer ball": 175,
701
+ "kissing": 176,
702
+ "kitesurfing": 177,
703
+ "knitting": 178,
704
+ "krumping": 179,
705
+ "laughing": 180,
706
+ "laying bricks": 181,
707
+ "long jump": 182,
708
+ "lunge": 183,
709
+ "making a cake": 184,
710
+ "making a sandwich": 185,
711
+ "making bed": 186,
712
+ "making jewelry": 187,
713
+ "making pizza": 188,
714
+ "making snowman": 189,
715
+ "making sushi": 190,
716
+ "making tea": 191,
717
+ "marching": 192,
718
+ "massaging back": 193,
719
+ "massaging feet": 194,
720
+ "massaging legs": 195,
721
+ "massaging person's head": 196,
722
+ "milking cow": 197,
723
+ "mopping floor": 198,
724
+ "motorcycling": 199,
725
+ "moving furniture": 200,
726
+ "mowing lawn": 201,
727
+ "news anchoring": 202,
728
+ "opening bottle": 203,
729
+ "opening present": 204,
730
+ "paragliding": 205,
731
+ "parasailing": 206,
732
+ "parkour": 207,
733
+ "passing American football (in game)": 208,
734
+ "passing American football (not in game)": 209,
735
+ "peeling apples": 210,
736
+ "peeling potatoes": 211,
737
+ "petting animal (not cat)": 212,
738
+ "petting cat": 213,
739
+ "picking fruit": 214,
740
+ "planting trees": 215,
741
+ "plastering": 216,
742
+ "playing accordion": 217,
743
+ "playing badminton": 218,
744
+ "playing bagpipes": 219,
745
+ "playing basketball": 220,
746
+ "playing bass guitar": 221,
747
+ "playing cards": 222,
748
+ "playing cello": 223,
749
+ "playing chess": 224,
750
+ "playing clarinet": 225,
751
+ "playing controller": 226,
752
+ "playing cricket": 227,
753
+ "playing cymbals": 228,
754
+ "playing didgeridoo": 229,
755
+ "playing drums": 230,
756
+ "playing flute": 231,
757
+ "playing guitar": 232,
758
+ "playing harmonica": 233,
759
+ "playing harp": 234,
760
+ "playing ice hockey": 235,
761
+ "playing keyboard": 236,
762
+ "playing kickball": 237,
763
+ "playing monopoly": 238,
764
+ "playing organ": 239,
765
+ "playing paintball": 240,
766
+ "playing piano": 241,
767
+ "playing poker": 242,
768
+ "playing recorder": 243,
769
+ "playing saxophone": 244,
770
+ "playing squash or racquetball": 245,
771
+ "playing tennis": 246,
772
+ "playing trombone": 247,
773
+ "playing trumpet": 248,
774
+ "playing ukulele": 249,
775
+ "playing violin": 250,
776
+ "playing volleyball": 251,
777
+ "playing xylophone": 252,
778
+ "pole vault": 253,
779
+ "presenting weather forecast": 254,
780
+ "pull ups": 255,
781
+ "pumping fist": 256,
782
+ "pumping gas": 257,
783
+ "punching bag": 258,
784
+ "punching person (boxing)": 259,
785
+ "push up": 260,
786
+ "pushing car": 261,
787
+ "pushing cart": 262,
788
+ "pushing wheelchair": 263,
789
+ "reading book": 264,
790
+ "reading newspaper": 265,
791
+ "recording music": 266,
792
+ "riding a bike": 267,
793
+ "riding camel": 268,
794
+ "riding elephant": 269,
795
+ "riding mechanical bull": 270,
796
+ "riding mountain bike": 271,
797
+ "riding mule": 272,
798
+ "riding or walking with horse": 273,
799
+ "riding scooter": 274,
800
+ "riding unicycle": 275,
801
+ "ripping paper": 276,
802
+ "robot dancing": 277,
803
+ "rock climbing": 278,
804
+ "rock scissors paper": 279,
805
+ "roller skating": 280,
806
+ "running on treadmill": 281,
807
+ "sailing": 282,
808
+ "salsa dancing": 283,
809
+ "sanding floor": 284,
810
+ "scrambling eggs": 285,
811
+ "scuba diving": 286,
812
+ "setting table": 287,
813
+ "shaking hands": 288,
814
+ "shaking head": 289,
815
+ "sharpening knives": 290,
816
+ "sharpening pencil": 291,
817
+ "shaving head": 292,
818
+ "shaving legs": 293,
819
+ "shearing sheep": 294,
820
+ "shining shoes": 295,
821
+ "shooting basketball": 296,
822
+ "shooting goal (soccer)": 297,
823
+ "shot put": 298,
824
+ "shoveling snow": 299,
825
+ "shredding paper": 300,
826
+ "shuffling cards": 301,
827
+ "side kick": 302,
828
+ "sign language interpreting": 303,
829
+ "singing": 304,
830
+ "situp": 305,
831
+ "skateboarding": 306,
832
+ "ski jumping": 307,
833
+ "skiing (not slalom or crosscountry)": 308,
834
+ "skiing crosscountry": 309,
835
+ "skiing slalom": 310,
836
+ "skipping rope": 311,
837
+ "skydiving": 312,
838
+ "slacklining": 313,
839
+ "slapping": 314,
840
+ "sled dog racing": 315,
841
+ "smoking": 316,
842
+ "smoking hookah": 317,
843
+ "snatch weight lifting": 318,
844
+ "sneezing": 319,
845
+ "sniffing": 320,
846
+ "snorkeling": 321,
847
+ "snowboarding": 322,
848
+ "snowkiting": 323,
849
+ "snowmobiling": 324,
850
+ "somersaulting": 325,
851
+ "spinning poi": 326,
852
+ "spray painting": 327,
853
+ "spraying": 328,
854
+ "springboard diving": 329,
855
+ "squat": 330,
856
+ "sticking tongue out": 331,
857
+ "stomping grapes": 332,
858
+ "stretching arm": 333,
859
+ "stretching leg": 334,
860
+ "strumming guitar": 335,
861
+ "surfing crowd": 336,
862
+ "surfing water": 337,
863
+ "sweeping floor": 338,
864
+ "swimming backstroke": 339,
865
+ "swimming breast stroke": 340,
866
+ "swimming butterfly stroke": 341,
867
+ "swing dancing": 342,
868
+ "swinging legs": 343,
869
+ "swinging on something": 344,
870
+ "sword fighting": 345,
871
+ "tai chi": 346,
872
+ "taking a shower": 347,
873
+ "tango dancing": 348,
874
+ "tap dancing": 349,
875
+ "tapping guitar": 350,
876
+ "tapping pen": 351,
877
+ "tasting beer": 352,
878
+ "tasting food": 353,
879
+ "testifying": 354,
880
+ "texting": 355,
881
+ "throwing axe": 356,
882
+ "throwing ball": 357,
883
+ "throwing discus": 358,
884
+ "tickling": 359,
885
+ "tobogganing": 360,
886
+ "tossing coin": 361,
887
+ "tossing salad": 362,
888
+ "training dog": 363,
889
+ "trapezing": 364,
890
+ "trimming or shaving beard": 365,
891
+ "trimming trees": 366,
892
+ "triple jump": 367,
893
+ "tying bow tie": 368,
894
+ "tying knot (not on a tie)": 369,
895
+ "tying tie": 370,
896
+ "unboxing": 371,
897
+ "unloading truck": 372,
898
+ "using computer": 373,
899
+ "using remote controller (not gaming)": 374,
900
+ "using segway": 375,
901
+ "vault": 376,
902
+ "waiting in line": 377,
903
+ "walking the dog": 378,
904
+ "washing dishes": 379,
905
+ "washing feet": 380,
906
+ "washing hair": 381,
907
+ "washing hands": 382,
908
+ "water skiing": 383,
909
+ "water sliding": 384,
910
+ "watering plants": 385,
911
+ "waxing back": 386,
912
+ "waxing chest": 387,
913
+ "waxing eyebrows": 388,
914
+ "waxing legs": 389,
915
+ "weaving basket": 390,
916
+ "welding": 391,
917
+ "whistling": 392,
918
+ "windsurfing": 393,
919
+ "wrapping present": 394,
920
+ "wrestling": 395,
921
+ "writing": 396,
922
+ "yawning": 397,
923
+ "yoga": 398,
924
+ "zumba": 399
925
+ },
926
+ "layer_norm_eps": 1e-06,
927
+ "length_penalty": 1.0,
928
+ "max_length": 20,
929
+ "min_length": 0,
930
+ "model_type": "timesformer",
931
+ "no_repeat_ngram_size": 0,
932
+ "num_attention_heads": 12,
933
+ "num_beam_groups": 1,
934
+ "num_beams": 1,
935
+ "num_channels": 3,
936
+ "num_frames": 8,
937
+ "num_hidden_layers": 12,
938
+ "num_return_sequences": 1,
939
+ "output_attentions": false,
940
+ "output_hidden_states": false,
941
+ "output_scores": false,
942
+ "pad_token_id": null,
943
+ "patch_size": 16,
944
+ "prefix": null,
945
+ "problem_type": null,
946
+ "pruned_heads": {},
947
+ "qkv_bias": true,
948
+ "remove_invalid_values": false,
949
+ "repetition_penalty": 1.0,
950
+ "return_dict": true,
951
+ "return_dict_in_generate": false,
952
+ "sep_token_id": null,
953
+ "suppress_tokens": null,
954
+ "task_specific_params": null,
955
+ "temperature": 1.0,
956
+ "tf_legacy_loss": false,
957
+ "tie_encoder_decoder": false,
958
+ "tie_word_embeddings": true,
959
+ "tokenizer_class": null,
960
+ "top_k": 50,
961
+ "top_p": 1.0,
962
+ "torch_dtype": "float32",
963
+ "torchscript": false,
964
+ "transformers_version": "4.30.2",
965
+ "typical_p": 1.0,
966
+ "use_bfloat16": false
967
+ },
968
+ "is_encoder_decoder": true,
969
+ "model_type": "vision-encoder-decoder",
970
+ "pad_token_id": 0,
971
+ "tie_word_embeddings": false,
972
+ "torch_dtype": "float32",
973
+ "transformers_version": null
974
+ }
generation_config.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "decoder_start_token_id": 101,
3
+ "pad_token_id": 0,
4
+ "transformers_version": "4.30.2"
5
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:550bbc864591948087fad67256537d7ba1c3f58ebdab57f05ee81cfc56fad4ae
3
+ size 1036755317