adirik commited on
Commit
5e9365c
1 Parent(s): 0c1ca22

Upload Mask2FormerForUniversalSegmentation

Browse files
Files changed (2) hide show
  1. config.json +314 -0
  2. pytorch_model.bin +3 -0
config.json ADDED
@@ -0,0 +1,314 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_commit_hash": null,
3
+ "activation_function": "relu",
4
+ "architectures": [
5
+ "Mask2FormerForUniversalSegmentation"
6
+ ],
7
+ "backbone_config": {
8
+ "_name_or_path": "",
9
+ "add_cross_attention": false,
10
+ "architectures": null,
11
+ "attention_probs_dropout_prob": 0.0,
12
+ "bad_words_ids": null,
13
+ "begin_suppress_tokens": null,
14
+ "bos_token_id": null,
15
+ "chunk_size_feed_forward": 0,
16
+ "cross_attention_hidden_size": null,
17
+ "decoder_start_token_id": null,
18
+ "depths": [
19
+ 2,
20
+ 2,
21
+ 18,
22
+ 2
23
+ ],
24
+ "diversity_penalty": 0.0,
25
+ "do_sample": false,
26
+ "drop_path_rate": 0.3,
27
+ "early_stopping": false,
28
+ "embed_dim": 128,
29
+ "encoder_no_repeat_ngram_size": 0,
30
+ "encoder_stride": 32,
31
+ "eos_token_id": null,
32
+ "exponential_decay_length_penalty": null,
33
+ "finetuning_task": null,
34
+ "forced_bos_token_id": null,
35
+ "forced_eos_token_id": null,
36
+ "hidden_act": "gelu",
37
+ "hidden_dropout_prob": 0.0,
38
+ "hidden_size": 1024,
39
+ "id2label": {
40
+ "0": "LABEL_0",
41
+ "1": "LABEL_1"
42
+ },
43
+ "image_size": 224,
44
+ "initializer_range": 0.02,
45
+ "is_decoder": false,
46
+ "is_encoder_decoder": false,
47
+ "label2id": {
48
+ "LABEL_0": 0,
49
+ "LABEL_1": 1
50
+ },
51
+ "layer_norm_eps": 1e-05,
52
+ "length_penalty": 1.0,
53
+ "max_length": 20,
54
+ "min_length": 0,
55
+ "mlp_ratio": 4.0,
56
+ "model_type": "swin",
57
+ "no_repeat_ngram_size": 0,
58
+ "num_beam_groups": 1,
59
+ "num_beams": 1,
60
+ "num_channels": 3,
61
+ "num_heads": [
62
+ 4,
63
+ 8,
64
+ 16,
65
+ 32
66
+ ],
67
+ "num_layers": 4,
68
+ "num_return_sequences": 1,
69
+ "out_features": [
70
+ "stage1",
71
+ "stage2",
72
+ "stage3",
73
+ "stage4"
74
+ ],
75
+ "output_attentions": false,
76
+ "output_hidden_states": false,
77
+ "output_scores": false,
78
+ "pad_token_id": null,
79
+ "patch_size": 4,
80
+ "path_norm": true,
81
+ "prefix": null,
82
+ "problem_type": null,
83
+ "pruned_heads": {},
84
+ "qkv_bias": true,
85
+ "remove_invalid_values": false,
86
+ "repetition_penalty": 1.0,
87
+ "return_dict": true,
88
+ "return_dict_in_generate": false,
89
+ "sep_token_id": null,
90
+ "stage_names": [
91
+ "stem",
92
+ "stage1",
93
+ "stage2",
94
+ "stage3",
95
+ "stage4"
96
+ ],
97
+ "suppress_tokens": null,
98
+ "task_specific_params": null,
99
+ "temperature": 1.0,
100
+ "tf_legacy_loss": false,
101
+ "tie_encoder_decoder": false,
102
+ "tie_word_embeddings": true,
103
+ "tokenizer_class": null,
104
+ "top_k": 50,
105
+ "top_p": 1.0,
106
+ "torch_dtype": null,
107
+ "torchscript": false,
108
+ "transformers_version": "4.26.0.dev0",
109
+ "typical_p": 1.0,
110
+ "use_absolute_embeddings": false,
111
+ "use_bfloat16": false,
112
+ "window_size": 12
113
+ },
114
+ "class_weight": 2.0,
115
+ "common_stride": 4,
116
+ "decoder_layers": 10,
117
+ "dice_weight": 5.0,
118
+ "dim_feedforward": 2048,
119
+ "dropout": 0.0,
120
+ "encoder_feedforward_dim": 1024,
121
+ "encoder_layers": 6,
122
+ "enforce_input_proj": false,
123
+ "enforce_input_projection": false,
124
+ "feature_size": 256,
125
+ "feature_strides": [
126
+ 4,
127
+ 8,
128
+ 16,
129
+ 32
130
+ ],
131
+ "hidden_dim": 256,
132
+ "id2label": {
133
+ "0": "person",
134
+ "1": "bicycle",
135
+ "2": "car",
136
+ "3": "motorbike",
137
+ "4": "aeroplane",
138
+ "5": "bus",
139
+ "6": "train",
140
+ "7": "truck",
141
+ "8": "boat",
142
+ "9": "traffic light",
143
+ "10": "fire hydrant",
144
+ "11": "stop sign",
145
+ "12": "parking meter",
146
+ "13": "bench",
147
+ "14": "bird",
148
+ "15": "cat",
149
+ "16": "dog",
150
+ "17": "horse",
151
+ "18": "sheep",
152
+ "19": "cow",
153
+ "20": "elephant",
154
+ "21": "bear",
155
+ "22": "zebra",
156
+ "23": "giraffe",
157
+ "24": "backpack",
158
+ "25": "umbrella",
159
+ "26": "handbag",
160
+ "27": "tie",
161
+ "28": "suitcase",
162
+ "29": "frisbee",
163
+ "30": "skis",
164
+ "31": "snowboard",
165
+ "32": "sports ball",
166
+ "33": "kite",
167
+ "34": "baseball bat",
168
+ "35": "baseball glove",
169
+ "36": "skateboard",
170
+ "37": "surfboard",
171
+ "38": "tennis racket",
172
+ "39": "bottle",
173
+ "40": "wine glass",
174
+ "41": "cup",
175
+ "42": "fork",
176
+ "43": "knife",
177
+ "44": "spoon",
178
+ "45": "bowl",
179
+ "46": "banana",
180
+ "47": "apple",
181
+ "48": "sandwich",
182
+ "49": "orange",
183
+ "50": "broccoli",
184
+ "51": "carrot",
185
+ "52": "hot dog",
186
+ "53": "pizza",
187
+ "54": "donut",
188
+ "55": "cake",
189
+ "56": "chair",
190
+ "57": "sofa",
191
+ "58": "pottedplant",
192
+ "59": "bed",
193
+ "60": "diningtable",
194
+ "61": "toilet",
195
+ "62": "tvmonitor",
196
+ "63": "laptop",
197
+ "64": "mouse",
198
+ "65": "remote",
199
+ "66": "keyboard",
200
+ "67": "cell phone",
201
+ "68": "microwave",
202
+ "69": "oven",
203
+ "70": "toaster",
204
+ "71": "sink",
205
+ "72": "refrigerator",
206
+ "73": "book",
207
+ "74": "clock",
208
+ "75": "vase",
209
+ "76": "scissors",
210
+ "77": "teddy bear",
211
+ "78": "hair drier",
212
+ "79": "toothbrush"
213
+ },
214
+ "ignore_value": 255,
215
+ "importance_sample_ratio": 0.75,
216
+ "init_std": 0.02,
217
+ "init_xavier_std": 1.0,
218
+ "label2id": {
219
+ "aeroplane": 4,
220
+ "apple": 47,
221
+ "backpack": 24,
222
+ "banana": 46,
223
+ "baseball bat": 34,
224
+ "baseball glove": 35,
225
+ "bear": 21,
226
+ "bed": 59,
227
+ "bench": 13,
228
+ "bicycle": 1,
229
+ "bird": 14,
230
+ "boat": 8,
231
+ "book": 73,
232
+ "bottle": 39,
233
+ "bowl": 45,
234
+ "broccoli": 50,
235
+ "bus": 5,
236
+ "cake": 55,
237
+ "car": 2,
238
+ "carrot": 51,
239
+ "cat": 15,
240
+ "cell phone": 67,
241
+ "chair": 56,
242
+ "clock": 74,
243
+ "cow": 19,
244
+ "cup": 41,
245
+ "diningtable": 60,
246
+ "dog": 16,
247
+ "donut": 54,
248
+ "elephant": 20,
249
+ "fire hydrant": 10,
250
+ "fork": 42,
251
+ "frisbee": 29,
252
+ "giraffe": 23,
253
+ "hair drier": 78,
254
+ "handbag": 26,
255
+ "horse": 17,
256
+ "hot dog": 52,
257
+ "keyboard": 66,
258
+ "kite": 33,
259
+ "knife": 43,
260
+ "laptop": 63,
261
+ "microwave": 68,
262
+ "motorbike": 3,
263
+ "mouse": 64,
264
+ "orange": 49,
265
+ "oven": 69,
266
+ "parking meter": 12,
267
+ "person": 0,
268
+ "pizza": 53,
269
+ "pottedplant": 58,
270
+ "refrigerator": 72,
271
+ "remote": 65,
272
+ "sandwich": 48,
273
+ "scissors": 76,
274
+ "sheep": 18,
275
+ "sink": 71,
276
+ "skateboard": 36,
277
+ "skis": 30,
278
+ "snowboard": 31,
279
+ "sofa": 57,
280
+ "spoon": 44,
281
+ "sports ball": 32,
282
+ "stop sign": 11,
283
+ "suitcase": 28,
284
+ "surfboard": 37,
285
+ "teddy bear": 77,
286
+ "tennis racket": 38,
287
+ "tie": 27,
288
+ "toaster": 70,
289
+ "toilet": 61,
290
+ "toothbrush": 79,
291
+ "traffic light": 9,
292
+ "train": 6,
293
+ "truck": 7,
294
+ "tvmonitor": 62,
295
+ "umbrella": 25,
296
+ "vase": 75,
297
+ "wine glass": 40,
298
+ "zebra": 22
299
+ },
300
+ "mask_feature_size": 256,
301
+ "mask_weight": 5.0,
302
+ "model_type": "mask2former",
303
+ "no_object_weight": 0.1,
304
+ "num_attention_heads": 8,
305
+ "num_hidden_layers": 10,
306
+ "num_queries": 100,
307
+ "output_auxiliary_logits": null,
308
+ "oversample_ratio": 3.0,
309
+ "pre_norm": false,
310
+ "torch_dtype": "float32",
311
+ "train_num_points": 12544,
312
+ "transformers_version": null,
313
+ "use_auxiliary_loss": true
314
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e4d54e1ec1ae5cb5fee47758bf47356b4b73b04a8b013abf48cb67a2fb1351a2
3
+ size 431878597