Upload config.json with huggingface_hub
Browse files- config.json +97 -0
config.json
ADDED
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "google/vit-base-patch16-224-in21k",
|
3 |
+
"architectures": [
|
4 |
+
"ViTModel"
|
5 |
+
],
|
6 |
+
"attention_probs_dropout_prob": 0.0,
|
7 |
+
"encoder_stride": 16,
|
8 |
+
"hidden_act": "gelu",
|
9 |
+
"hidden_dropout_prob": 0.0,
|
10 |
+
"hidden_size": 768,
|
11 |
+
"id2label": {
|
12 |
+
"0": "apartment_building",
|
13 |
+
"1": "basketball_court",
|
14 |
+
"2": "bathroom",
|
15 |
+
"3": "beach",
|
16 |
+
"4": "bedroom",
|
17 |
+
"5": "classroom",
|
18 |
+
"6": "field",
|
19 |
+
"7": "forest",
|
20 |
+
"8": "highway",
|
21 |
+
"9": "house",
|
22 |
+
"10": "industrial_area",
|
23 |
+
"11": "kitchen",
|
24 |
+
"12": "lake",
|
25 |
+
"13": "library",
|
26 |
+
"14": "lobby",
|
27 |
+
"15": "martial_arts_gym",
|
28 |
+
"16": "mountain",
|
29 |
+
"17": "movie_theater",
|
30 |
+
"18": "music_studio",
|
31 |
+
"19": "office",
|
32 |
+
"20": "park",
|
33 |
+
"21": "playground",
|
34 |
+
"22": "pub",
|
35 |
+
"23": "restaurant",
|
36 |
+
"24": "shopfront",
|
37 |
+
"25": "street",
|
38 |
+
"26": "subway_station",
|
39 |
+
"27": "supermarket",
|
40 |
+
"28": "swimming_pool",
|
41 |
+
"29": "train_station",
|
42 |
+
"30": "village",
|
43 |
+
"31": "volcano",
|
44 |
+
"32": "waiting_room",
|
45 |
+
"33": "wind_farm",
|
46 |
+
"34": "yard"
|
47 |
+
},
|
48 |
+
"image_size": 224,
|
49 |
+
"initializer_range": 0.02,
|
50 |
+
"intermediate_size": 3072,
|
51 |
+
"label2id": {
|
52 |
+
"apartment_building": 0,
|
53 |
+
"basketball_court": 1,
|
54 |
+
"bathroom": 2,
|
55 |
+
"beach": 3,
|
56 |
+
"bedroom": 4,
|
57 |
+
"classroom": 5,
|
58 |
+
"field": 6,
|
59 |
+
"forest": 7,
|
60 |
+
"highway": 8,
|
61 |
+
"house": 9,
|
62 |
+
"industrial_area": 10,
|
63 |
+
"kitchen": 11,
|
64 |
+
"lake": 12,
|
65 |
+
"library": 13,
|
66 |
+
"lobby": 14,
|
67 |
+
"martial_arts_gym": 15,
|
68 |
+
"mountain": 16,
|
69 |
+
"movie_theater": 17,
|
70 |
+
"music_studio": 18,
|
71 |
+
"office": 19,
|
72 |
+
"park": 20,
|
73 |
+
"playground": 21,
|
74 |
+
"pub": 22,
|
75 |
+
"restaurant": 23,
|
76 |
+
"shopfront": 24,
|
77 |
+
"street": 25,
|
78 |
+
"subway_station": 26,
|
79 |
+
"supermarket": 27,
|
80 |
+
"swimming_pool": 28,
|
81 |
+
"train_station": 29,
|
82 |
+
"village": 30,
|
83 |
+
"volcano": 31,
|
84 |
+
"waiting_room": 32,
|
85 |
+
"wind_farm": 33,
|
86 |
+
"yard": 34
|
87 |
+
},
|
88 |
+
"layer_norm_eps": 1e-12,
|
89 |
+
"model_type": "vit",
|
90 |
+
"num_attention_heads": 12,
|
91 |
+
"num_channels": 3,
|
92 |
+
"num_hidden_layers": 12,
|
93 |
+
"patch_size": 16,
|
94 |
+
"qkv_bias": true,
|
95 |
+
"torch_dtype": "float32",
|
96 |
+
"transformers_version": "4.46.0"
|
97 |
+
}
|