praeclarumjj3 commited on
Commit
d962224
1 Parent(s): da593c5
Files changed (2) hide show
  1. config.json +412 -0
  2. pytorch_model.bin +3 -0
config.json ADDED
@@ -0,0 +1,412 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "OneFormerForUniversalSegmentation"
4
+ ],
5
+ "backbone_config": {
6
+ "attention_probs_dropout_prob": 0.0,
7
+ "depths": [
8
+ 3,
9
+ 4,
10
+ 18,
11
+ 5
12
+ ],
13
+ "dilations": [
14
+ [
15
+ 1,
16
+ 20,
17
+ 1
18
+ ],
19
+ [
20
+ 1,
21
+ 5,
22
+ 1,
23
+ 10
24
+ ],
25
+ [
26
+ 1,
27
+ 2,
28
+ 1,
29
+ 3,
30
+ 1,
31
+ 4,
32
+ 1,
33
+ 5,
34
+ 1,
35
+ 2,
36
+ 1,
37
+ 3,
38
+ 1,
39
+ 4,
40
+ 1,
41
+ 5,
42
+ 1,
43
+ 5
44
+ ],
45
+ [
46
+ 1,
47
+ 2,
48
+ 1,
49
+ 2,
50
+ 1
51
+ ]
52
+ ],
53
+ "drop_path_rate": 0.3,
54
+ "embed_dim": 192,
55
+ "encoder_stride": 32,
56
+ "feature_channels": [
57
+ 192,
58
+ 384,
59
+ 768,
60
+ 1536
61
+ ],
62
+ "hidden_act": "gelu",
63
+ "hidden_dropout_prob": 0,
64
+ "kernel_size": 11,
65
+ "layer_scale_init_value": 0.0,
66
+ "mlp_ratio": 2.0,
67
+ "num_channels": 3,
68
+ "num_heads": [
69
+ 6,
70
+ 12,
71
+ 24,
72
+ 48
73
+ ],
74
+ "patch_size": 4,
75
+ "qkv_bias": true,
76
+ "strides": [
77
+ 4,
78
+ 8,
79
+ 16,
80
+ 32
81
+ ]
82
+ },
83
+ "decoder_config": {
84
+ "common_stride": 4,
85
+ "conv_dim": 256,
86
+ "decoder_layers": 10,
87
+ "dim_feedforward": 2048,
88
+ "dropout": 0.1,
89
+ "encoder_feedforward_dim": 1024,
90
+ "encoder_layers": 6,
91
+ "enforce_input_proj": false,
92
+ "hidden_dim": 256,
93
+ "mask_dim": 256,
94
+ "norm": "GN",
95
+ "num_heads": 8,
96
+ "pre_norm": false,
97
+ "query_dec_layers": 2,
98
+ "use_task_norm": true
99
+ },
100
+ "general_config": {
101
+ "backbone_type": "dinat",
102
+ "class_weight": 2.0,
103
+ "contrastive_temperature": 0.07,
104
+ "contrastive_weight": 0.5,
105
+ "deep_supervision": true,
106
+ "dice_weight": 5.0,
107
+ "ignore_value": 255,
108
+ "importance_sample_ratio": 0.75,
109
+ "init_std": 0.02,
110
+ "init_xavier_std": 1.0,
111
+ "is_train": false,
112
+ "layer_norm_eps": 1e-05,
113
+ "mask_weight": 5.0,
114
+ "no_object_weight": 0.1,
115
+ "num_classes": 133,
116
+ "num_queries": 150,
117
+ "output_auxiliary_logits": true,
118
+ "oversample_ratio": 3.0,
119
+ "train_num_points": 12544,
120
+ "use_auxiliary_loss": true
121
+ },
122
+ "hidden_size": 256,
123
+ "id2label": {
124
+ "0": "person",
125
+ "1": "bicycle",
126
+ "2": "car",
127
+ "3": "motorcycle",
128
+ "4": "airplane",
129
+ "5": "bus",
130
+ "6": "train",
131
+ "7": "truck",
132
+ "8": "boat",
133
+ "9": "traffic light",
134
+ "10": "fire hydrant",
135
+ "11": "stop sign",
136
+ "12": "parking meter",
137
+ "13": "bench",
138
+ "14": "bird",
139
+ "15": "cat",
140
+ "16": "dog",
141
+ "17": "horse",
142
+ "18": "sheep",
143
+ "19": "cow",
144
+ "20": "elephant",
145
+ "21": "bear",
146
+ "22": "zebra",
147
+ "23": "giraffe",
148
+ "24": "backpack",
149
+ "25": "umbrella",
150
+ "26": "handbag",
151
+ "27": "tie",
152
+ "28": "suitcase",
153
+ "29": "frisbee",
154
+ "30": "skis",
155
+ "31": "snowboard",
156
+ "32": "sports ball",
157
+ "33": "kite",
158
+ "34": "baseball bat",
159
+ "35": "baseball glove",
160
+ "36": "skateboard",
161
+ "37": "surfboard",
162
+ "38": "tennis racket",
163
+ "39": "bottle",
164
+ "40": "wine glass",
165
+ "41": "cup",
166
+ "42": "fork",
167
+ "43": "knife",
168
+ "44": "spoon",
169
+ "45": "bowl",
170
+ "46": "banana",
171
+ "47": "apple",
172
+ "48": "sandwich",
173
+ "49": "orange",
174
+ "50": "broccoli",
175
+ "51": "carrot",
176
+ "52": "hot dog",
177
+ "53": "pizza",
178
+ "54": "donut",
179
+ "55": "cake",
180
+ "56": "chair",
181
+ "57": "couch",
182
+ "58": "potted plant",
183
+ "59": "bed",
184
+ "60": "dining table",
185
+ "61": "toilet",
186
+ "62": "tv",
187
+ "63": "laptop",
188
+ "64": "mouse",
189
+ "65": "remote",
190
+ "66": "keyboard",
191
+ "67": "cell phone",
192
+ "68": "microwave",
193
+ "69": "oven",
194
+ "70": "toaster",
195
+ "71": "sink",
196
+ "72": "refrigerator",
197
+ "73": "book",
198
+ "74": "clock",
199
+ "75": "vase",
200
+ "76": "scissors",
201
+ "77": "teddy bear",
202
+ "78": "hair drier",
203
+ "79": "toothbrush",
204
+ "80": "banner",
205
+ "81": "blanket",
206
+ "82": "bridge",
207
+ "83": "cardboard",
208
+ "84": "counter",
209
+ "85": "curtain",
210
+ "86": "door-stuff",
211
+ "87": "floor-wood",
212
+ "88": "flower",
213
+ "89": "fruit",
214
+ "90": "gravel",
215
+ "91": "house",
216
+ "92": "light",
217
+ "93": "mirror-stuff",
218
+ "94": "net",
219
+ "95": "pillow",
220
+ "96": "platform",
221
+ "97": "playingfield",
222
+ "98": "railroad",
223
+ "99": "river",
224
+ "100": "road",
225
+ "101": "roof",
226
+ "102": "sand",
227
+ "103": "sea",
228
+ "104": "shelf",
229
+ "105": "snow",
230
+ "106": "stairs",
231
+ "107": "tent",
232
+ "108": "towel",
233
+ "109": "wall-brick",
234
+ "110": "wall-stone",
235
+ "111": "wall-tile",
236
+ "112": "wall-wood",
237
+ "113": "water-other",
238
+ "114": "window-blind",
239
+ "115": "window-other",
240
+ "116": "tree-merged",
241
+ "117": "fence-merged",
242
+ "118": "ceiling-merged",
243
+ "119": "sky-other-merged",
244
+ "120": "cabinet-merged",
245
+ "121": "table-merged",
246
+ "122": "floor-other-merged",
247
+ "123": "pavement-merged",
248
+ "124": "mountain-merged",
249
+ "125": "grass-merged",
250
+ "126": "dirt-merged",
251
+ "127": "paper-merged",
252
+ "128": "food-other-merged",
253
+ "129": "building-other-merged",
254
+ "130": "rock-merged",
255
+ "131": "wall-other-merged",
256
+ "132": "rug-merged"
257
+ },
258
+ "init_std": 0.02,
259
+ "init_xavier_std": 1.0,
260
+ "label2id": {
261
+ "airplane": 4,
262
+ "apple": 47,
263
+ "backpack": 24,
264
+ "banana": 46,
265
+ "banner": 80,
266
+ "baseball bat": 34,
267
+ "baseball glove": 35,
268
+ "bear": 21,
269
+ "bed": 59,
270
+ "bench": 13,
271
+ "bicycle": 1,
272
+ "bird": 14,
273
+ "blanket": 81,
274
+ "boat": 8,
275
+ "book": 73,
276
+ "bottle": 39,
277
+ "bowl": 45,
278
+ "bridge": 82,
279
+ "broccoli": 50,
280
+ "building-other-merged": 129,
281
+ "bus": 5,
282
+ "cabinet-merged": 120,
283
+ "cake": 55,
284
+ "car": 2,
285
+ "cardboard": 83,
286
+ "carrot": 51,
287
+ "cat": 15,
288
+ "ceiling-merged": 118,
289
+ "cell phone": 67,
290
+ "chair": 56,
291
+ "clock": 74,
292
+ "couch": 57,
293
+ "counter": 84,
294
+ "cow": 19,
295
+ "cup": 41,
296
+ "curtain": 85,
297
+ "dining table": 60,
298
+ "dirt-merged": 126,
299
+ "dog": 16,
300
+ "donut": 54,
301
+ "door-stuff": 86,
302
+ "elephant": 20,
303
+ "fence-merged": 117,
304
+ "fire hydrant": 10,
305
+ "floor-other-merged": 122,
306
+ "floor-wood": 87,
307
+ "flower": 88,
308
+ "food-other-merged": 128,
309
+ "fork": 42,
310
+ "frisbee": 29,
311
+ "fruit": 89,
312
+ "giraffe": 23,
313
+ "grass-merged": 125,
314
+ "gravel": 90,
315
+ "hair drier": 78,
316
+ "handbag": 26,
317
+ "horse": 17,
318
+ "hot dog": 52,
319
+ "house": 91,
320
+ "keyboard": 66,
321
+ "kite": 33,
322
+ "knife": 43,
323
+ "laptop": 63,
324
+ "light": 92,
325
+ "microwave": 68,
326
+ "mirror-stuff": 93,
327
+ "motorcycle": 3,
328
+ "mountain-merged": 124,
329
+ "mouse": 64,
330
+ "net": 94,
331
+ "orange": 49,
332
+ "oven": 69,
333
+ "paper-merged": 127,
334
+ "parking meter": 12,
335
+ "pavement-merged": 123,
336
+ "person": 0,
337
+ "pillow": 95,
338
+ "pizza": 53,
339
+ "platform": 96,
340
+ "playingfield": 97,
341
+ "potted plant": 58,
342
+ "railroad": 98,
343
+ "refrigerator": 72,
344
+ "remote": 65,
345
+ "river": 99,
346
+ "road": 100,
347
+ "rock-merged": 130,
348
+ "roof": 101,
349
+ "rug-merged": 132,
350
+ "sand": 102,
351
+ "sandwich": 48,
352
+ "scissors": 76,
353
+ "sea": 103,
354
+ "sheep": 18,
355
+ "shelf": 104,
356
+ "sink": 71,
357
+ "skateboard": 36,
358
+ "skis": 30,
359
+ "sky-other-merged": 119,
360
+ "snow": 105,
361
+ "snowboard": 31,
362
+ "spoon": 44,
363
+ "sports ball": 32,
364
+ "stairs": 106,
365
+ "stop sign": 11,
366
+ "suitcase": 28,
367
+ "surfboard": 37,
368
+ "table-merged": 121,
369
+ "teddy bear": 77,
370
+ "tennis racket": 38,
371
+ "tent": 107,
372
+ "tie": 27,
373
+ "toaster": 70,
374
+ "toilet": 61,
375
+ "toothbrush": 79,
376
+ "towel": 108,
377
+ "traffic light": 9,
378
+ "train": 6,
379
+ "tree-merged": 116,
380
+ "truck": 7,
381
+ "tv": 62,
382
+ "umbrella": 25,
383
+ "vase": 75,
384
+ "wall-brick": 109,
385
+ "wall-other-merged": 131,
386
+ "wall-stone": 110,
387
+ "wall-tile": 111,
388
+ "wall-wood": 112,
389
+ "water-other": 113,
390
+ "window-blind": 114,
391
+ "window-other": 115,
392
+ "wine glass": 40,
393
+ "zebra": 22
394
+ },
395
+ "model_type": "oneformer",
396
+ "num_attention_heads": 8,
397
+ "num_hidden_layers": 10,
398
+ "output_attentions": true,
399
+ "output_hidden_states": true,
400
+ "text_encoder_config": {
401
+ "max_seq_len": 77,
402
+ "task_seq_len": 77,
403
+ "text_encoder_context_length": 77,
404
+ "text_encoder_n_ctx": 16,
405
+ "text_encoder_num_layers": 6,
406
+ "text_encoder_proj_layers": 2,
407
+ "text_encoder_vocab_size": 49408,
408
+ "text_encoder_width": 256
409
+ },
410
+ "torch_dtype": "float32",
411
+ "transformers_version": "4.25.0.dev0"
412
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:edd2b4d02b8b498af82bab75dbc7110392d1d5b7573820440e2bc2d34a65b53a
3
+ size 893082029