config.json
14.6 KB · 388 lines · json Raw
1 {
2 "architectures": [
3 "VJEPA2ForVideoClassification"
4 ],
5 "attention_dropout": 0.0,
6 "attention_probs_dropout_prob": 0.0,
7 "crop_size": 256,
8 "drop_path_rate": 0.0,
9 "frames_per_clip": 16,
10 "hidden_act": "gelu",
11 "hidden_dropout_prob": 0.0,
12 "hidden_size": 1024,
13 "id2label": {
14 "0": "Approaching [something] with your camera",
15 "1": "Attaching [something] to [something]",
16 "2": "Bending [something] so that it deforms",
17 "3": "Bending [something] until it breaks",
18 "4": "Burying [something] in [something]",
19 "5": "Closing [something]",
20 "6": "Covering [something] with [something]",
21 "7": "Digging [something] out of [something]",
22 "8": "Dropping [something] behind [something]",
23 "9": "Dropping [something] in front of [something]",
24 "10": "Dropping [something] into [something]",
25 "11": "Dropping [something] next to [something]",
26 "12": "Dropping [something] onto [something]",
27 "13": "Failing to put [something] into [something] because [something] does not fit",
28 "14": "Folding [something]",
29 "15": "Hitting [something] with [something]",
30 "16": "Holding [something]",
31 "17": "Holding [something] behind [something]",
32 "18": "Holding [something] in front of [something]",
33 "19": "Holding [something] next to [something]",
34 "20": "Holding [something] over [something]",
35 "21": "Laying [something] on the table on its side, not upright",
36 "22": "Letting [something] roll along a flat surface",
37 "23": "Letting [something] roll down a slanted surface",
38 "24": "Letting [something] roll up a slanted surface, so it rolls back down",
39 "25": "Lifting a surface with [something] on it but not enough for it to slide down",
40 "26": "Lifting a surface with [something] on it until it starts sliding down",
41 "27": "Lifting [something] up completely without letting it drop down",
42 "28": "Lifting [something] up completely, then letting it drop down",
43 "29": "Lifting [something] with [something] on it",
44 "30": "Lifting up one end of [something] without letting it drop down",
45 "31": "Lifting up one end of [something], then letting it drop down",
46 "32": "Moving away from [something] with your camera",
47 "33": "Moving [part] of [something]",
48 "34": "Moving [something] across a surface until it falls down",
49 "35": "Moving [something] across a surface without it falling down",
50 "36": "Moving [something] and [something] away from each other",
51 "37": "Moving [something] and [something] closer to each other",
52 "38": "Moving [something] and [something] so they collide with each other",
53 "39": "Moving [something] and [something] so they pass each other",
54 "40": "Moving [something] away from [something]",
55 "41": "Moving [something] away from the camera",
56 "42": "Moving [something] closer to [something]",
57 "43": "Moving [something] down",
58 "44": "Moving [something] towards the camera",
59 "45": "Moving [something] up",
60 "46": "Opening [something]",
61 "47": "Picking [something] up",
62 "48": "Piling [something] up",
63 "49": "Plugging [something] into [something]",
64 "50": "Plugging [something] into [something] but pulling it right out as you remove your hand",
65 "51": "Poking a hole into [some substance]",
66 "52": "Poking a hole into [something soft]",
67 "53": "Poking a stack of [something] so the stack collapses",
68 "54": "Poking a stack of [something] without the stack collapsing",
69 "55": "Poking [something] so it slightly moves",
70 "56": "Poking [something] so lightly that it doesn't or almost doesn't move",
71 "57": "Poking [something] so that it falls over",
72 "58": "Poking [something] so that it spins around",
73 "59": "Pouring [something] into [something]",
74 "60": "Pouring [something] into [something] until it overflows",
75 "61": "Pouring [something] onto [something]",
76 "62": "Pouring [something] out of [something]",
77 "63": "Pretending or failing to wipe [something] off of [something]",
78 "64": "Pretending or trying and failing to twist [something]",
79 "65": "Pretending to be tearing [something that is not tearable]",
80 "66": "Pretending to close [something] without actually closing it",
81 "67": "Pretending to open [something] without actually opening it",
82 "68": "Pretending to pick [something] up",
83 "69": "Pretending to poke [something]",
84 "70": "Pretending to pour [something] out of [something], but [something] is empty",
85 "71": "Pretending to put [something] behind [something]",
86 "72": "Pretending to put [something] into [something]",
87 "73": "Pretending to put [something] next to [something]",
88 "74": "Pretending to put [something] on a surface",
89 "75": "Pretending to put [something] onto [something]",
90 "76": "Pretending to put [something] underneath [something]",
91 "77": "Pretending to scoop [something] up with [something]",
92 "78": "Pretending to spread air onto [something]",
93 "79": "Pretending to sprinkle air onto [something]",
94 "80": "Pretending to squeeze [something]",
95 "81": "Pretending to take [something] from [somewhere]",
96 "82": "Pretending to take [something] out of [something]",
97 "83": "Pretending to throw [something]",
98 "84": "Pretending to turn [something] upside down",
99 "85": "Pulling [something] from behind of [something]",
100 "86": "Pulling [something] from left to right",
101 "87": "Pulling [something] from right to left",
102 "88": "Pulling [something] onto [something]",
103 "89": "Pulling [something] out of [something]",
104 "90": "Pulling two ends of [something] but nothing happens",
105 "91": "Pulling two ends of [something] so that it gets stretched",
106 "92": "Pulling two ends of [something] so that it separates into two pieces",
107 "93": "Pushing [something] from left to right",
108 "94": "Pushing [something] from right to left",
109 "95": "Pushing [something] off of [something]",
110 "96": "Pushing [something] onto [something]",
111 "97": "Pushing [something] so it spins",
112 "98": "Pushing [something] so that it almost falls off but doesn't",
113 "99": "Pushing [something] so that it falls off the table",
114 "100": "Pushing [something] so that it slightly moves",
115 "101": "Pushing [something] with [something]",
116 "102": "Putting [number of] [something] onto [something]",
117 "103": "Putting [something] and [something] on the table",
118 "104": "Putting [something] behind [something]",
119 "105": "Putting [something] in front of [something]",
120 "106": "Putting [something] into [something]",
121 "107": "Putting [something] next to [something]",
122 "108": "Putting [something] on a flat surface without letting it roll",
123 "109": "Putting [something] on a surface",
124 "110": "Putting [something] on the edge of [something] so it is not supported and falls down",
125 "111": "Putting [something] onto a slanted surface but it doesn't glide down",
126 "112": "Putting [something] onto [something]",
127 "113": "Putting [something] onto [something else that cannot support it] so it falls down",
128 "114": "Putting [something similar to other things that are already on the table]",
129 "115": "Putting [something] that can't roll onto a slanted surface, so it slides down",
130 "116": "Putting [something] that can't roll onto a slanted surface, so it stays where it is",
131 "117": "Putting [something that cannot actually stand upright] upright on the table, so it falls on its side",
132 "118": "Putting [something] underneath [something]",
133 "119": "Putting [something] upright on the table",
134 "120": "Putting [something], [something] and [something] on the table",
135 "121": "Removing [something], revealing [something] behind",
136 "122": "Rolling [something] on a flat surface",
137 "123": "Scooping [something] up with [something]",
138 "124": "Showing a photo of [something] to the camera",
139 "125": "Showing [something] behind [something]",
140 "126": "Showing [something] next to [something]",
141 "127": "Showing [something] on top of [something]",
142 "128": "Showing [something] to the camera",
143 "129": "Showing that [something] is empty",
144 "130": "Showing that [something] is inside [something]",
145 "131": "[Something] being deflected from [something]",
146 "132": "[Something] colliding with [something] and both are being deflected",
147 "133": "[Something] colliding with [something] and both come to a halt",
148 "134": "[Something] falling like a feather or paper",
149 "135": "[Something] falling like a rock",
150 "136": "Spilling [something] behind [something]",
151 "137": "Spilling [something] next to [something]",
152 "138": "Spilling [something] onto [something]",
153 "139": "Spinning [something] so it continues spinning",
154 "140": "Spinning [something] that quickly stops spinning",
155 "141": "Spreading [something] onto [something]",
156 "142": "Sprinkling [something] onto [something]",
157 "143": "Squeezing [something]",
158 "144": "Stacking [number of] [something]",
159 "145": "Stuffing [something] into [something]",
160 "146": "Taking [one of many similar things on the table]",
161 "147": "Taking [something] from [somewhere]",
162 "148": "Taking [something] out of [something]",
163 "149": "Tearing [something] into two pieces",
164 "150": "Tearing [something] just a little bit",
165 "151": "Throwing [something]",
166 "152": "Throwing [something] against [something]",
167 "153": "Throwing [something] in the air and catching it",
168 "154": "Throwing [something] in the air and letting it fall",
169 "155": "Throwing [something] onto a surface",
170 "156": "Tilting [something] with [something] on it slightly so it doesn't fall down",
171 "157": "Tilting [something] with [something] on it until it falls off",
172 "158": "Tipping [something] over",
173 "159": "Tipping [something] with [something in it] over, so [something in it] falls out",
174 "160": "Touching (without moving) [part] of [something]",
175 "161": "Trying but failing to attach [something] to [something] because it doesn't stick",
176 "162": "Trying to bend [something unbendable] so nothing happens",
177 "163": "Trying to pour [something] into [something], but missing so it spills next to it",
178 "164": "Turning [something] upside down",
179 "165": "Turning the camera downwards while filming [something]",
180 "166": "Turning the camera left while filming [something]",
181 "167": "Turning the camera right while filming [something]",
182 "168": "Turning the camera upwards while filming [something]",
183 "169": "Twisting (wringing) [something] wet until water comes out",
184 "170": "Twisting [something]",
185 "171": "Uncovering [something]",
186 "172": "Unfolding [something]",
187 "173": "Wiping [something] off of [something]"
188 },
189 "image_size": 256,
190 "in_chans": 3,
191 "initializer_range": 0.02,
192 "label2id": {
193 "LABEL_0": 0,
194 "LABEL_1": 1,
195 "LABEL_10": 10,
196 "LABEL_100": 100,
197 "LABEL_101": 101,
198 "LABEL_102": 102,
199 "LABEL_103": 103,
200 "LABEL_104": 104,
201 "LABEL_105": 105,
202 "LABEL_106": 106,
203 "LABEL_107": 107,
204 "LABEL_108": 108,
205 "LABEL_109": 109,
206 "LABEL_11": 11,
207 "LABEL_110": 110,
208 "LABEL_111": 111,
209 "LABEL_112": 112,
210 "LABEL_113": 113,
211 "LABEL_114": 114,
212 "LABEL_115": 115,
213 "LABEL_116": 116,
214 "LABEL_117": 117,
215 "LABEL_118": 118,
216 "LABEL_119": 119,
217 "LABEL_12": 12,
218 "LABEL_120": 120,
219 "LABEL_121": 121,
220 "LABEL_122": 122,
221 "LABEL_123": 123,
222 "LABEL_124": 124,
223 "LABEL_125": 125,
224 "LABEL_126": 126,
225 "LABEL_127": 127,
226 "LABEL_128": 128,
227 "LABEL_129": 129,
228 "LABEL_13": 13,
229 "LABEL_130": 130,
230 "LABEL_131": 131,
231 "LABEL_132": 132,
232 "LABEL_133": 133,
233 "LABEL_134": 134,
234 "LABEL_135": 135,
235 "LABEL_136": 136,
236 "LABEL_137": 137,
237 "LABEL_138": 138,
238 "LABEL_139": 139,
239 "LABEL_14": 14,
240 "LABEL_140": 140,
241 "LABEL_141": 141,
242 "LABEL_142": 142,
243 "LABEL_143": 143,
244 "LABEL_144": 144,
245 "LABEL_145": 145,
246 "LABEL_146": 146,
247 "LABEL_147": 147,
248 "LABEL_148": 148,
249 "LABEL_149": 149,
250 "LABEL_15": 15,
251 "LABEL_150": 150,
252 "LABEL_151": 151,
253 "LABEL_152": 152,
254 "LABEL_153": 153,
255 "LABEL_154": 154,
256 "LABEL_155": 155,
257 "LABEL_156": 156,
258 "LABEL_157": 157,
259 "LABEL_158": 158,
260 "LABEL_159": 159,
261 "LABEL_16": 16,
262 "LABEL_160": 160,
263 "LABEL_161": 161,
264 "LABEL_162": 162,
265 "LABEL_163": 163,
266 "LABEL_164": 164,
267 "LABEL_165": 165,
268 "LABEL_166": 166,
269 "LABEL_167": 167,
270 "LABEL_168": 168,
271 "LABEL_169": 169,
272 "LABEL_17": 17,
273 "LABEL_170": 170,
274 "LABEL_171": 171,
275 "LABEL_172": 172,
276 "LABEL_173": 173,
277 "LABEL_18": 18,
278 "LABEL_19": 19,
279 "LABEL_2": 2,
280 "LABEL_20": 20,
281 "LABEL_21": 21,
282 "LABEL_22": 22,
283 "LABEL_23": 23,
284 "LABEL_24": 24,
285 "LABEL_25": 25,
286 "LABEL_26": 26,
287 "LABEL_27": 27,
288 "LABEL_28": 28,
289 "LABEL_29": 29,
290 "LABEL_3": 3,
291 "LABEL_30": 30,
292 "LABEL_31": 31,
293 "LABEL_32": 32,
294 "LABEL_33": 33,
295 "LABEL_34": 34,
296 "LABEL_35": 35,
297 "LABEL_36": 36,
298 "LABEL_37": 37,
299 "LABEL_38": 38,
300 "LABEL_39": 39,
301 "LABEL_4": 4,
302 "LABEL_40": 40,
303 "LABEL_41": 41,
304 "LABEL_42": 42,
305 "LABEL_43": 43,
306 "LABEL_44": 44,
307 "LABEL_45": 45,
308 "LABEL_46": 46,
309 "LABEL_47": 47,
310 "LABEL_48": 48,
311 "LABEL_49": 49,
312 "LABEL_5": 5,
313 "LABEL_50": 50,
314 "LABEL_51": 51,
315 "LABEL_52": 52,
316 "LABEL_53": 53,
317 "LABEL_54": 54,
318 "LABEL_55": 55,
319 "LABEL_56": 56,
320 "LABEL_57": 57,
321 "LABEL_58": 58,
322 "LABEL_59": 59,
323 "LABEL_6": 6,
324 "LABEL_60": 60,
325 "LABEL_61": 61,
326 "LABEL_62": 62,
327 "LABEL_63": 63,
328 "LABEL_64": 64,
329 "LABEL_65": 65,
330 "LABEL_66": 66,
331 "LABEL_67": 67,
332 "LABEL_68": 68,
333 "LABEL_69": 69,
334 "LABEL_7": 7,
335 "LABEL_70": 70,
336 "LABEL_71": 71,
337 "LABEL_72": 72,
338 "LABEL_73": 73,
339 "LABEL_74": 74,
340 "LABEL_75": 75,
341 "LABEL_76": 76,
342 "LABEL_77": 77,
343 "LABEL_78": 78,
344 "LABEL_79": 79,
345 "LABEL_8": 8,
346 "LABEL_80": 80,
347 "LABEL_81": 81,
348 "LABEL_82": 82,
349 "LABEL_83": 83,
350 "LABEL_84": 84,
351 "LABEL_85": 85,
352 "LABEL_86": 86,
353 "LABEL_87": 87,
354 "LABEL_88": 88,
355 "LABEL_89": 89,
356 "LABEL_9": 9,
357 "LABEL_90": 90,
358 "LABEL_91": 91,
359 "LABEL_92": 92,
360 "LABEL_93": 93,
361 "LABEL_94": 94,
362 "LABEL_95": 95,
363 "LABEL_96": 96,
364 "LABEL_97": 97,
365 "LABEL_98": 98,
366 "LABEL_99": 99
367 },
368 "layer_norm_eps": 1e-06,
369 "mlp_ratio": 4,
370 "model_type": "vjepa2",
371 "num_attention_heads": 16,
372 "num_hidden_layers": 24,
373 "num_pooler_layers": 3,
374 "patch_size": 16,
375 "pred_hidden_size": 384,
376 "pred_mlp_ratio": 4.0,
377 "pred_num_attention_heads": 12,
378 "pred_num_hidden_layers": 12,
379 "pred_num_mask_tokens": 10,
380 "pred_zero_init_mask_tokens": true,
381 "qkv_bias": true,
382 "torch_dtype": "float32",
383 "transformers_version": "4.53.0.dev0",
384 "tubelet_size": 2,
385 "use_SiLU": false,
386 "wide_SiLU": true
387 }
388