config.json
2.2 KB · 115 lines · json Raw
1 {
2 "add_projection": false,
3 "architectures": [
4 "ZoeDepthForDepthEstimation"
5 ],
6 "attractor_alpha": 1000,
7 "attractor_gamma": 2,
8 "attractor_kind": "mean",
9 "backbone": null,
10 "backbone_config": {
11 "hidden_size": 1024,
12 "image_size": 384,
13 "intermediate_size": 4096,
14 "model_type": "beit",
15 "num_attention_heads": 16,
16 "num_hidden_layers": 24,
17 "out_features": [
18 "stage6",
19 "stage12",
20 "stage18",
21 "stage24"
22 ],
23 "out_indices": [
24 6,
25 12,
26 18,
27 24
28 ],
29 "reshape_hidden_states": false,
30 "stage_names": [
31 "stem",
32 "stage1",
33 "stage2",
34 "stage3",
35 "stage4",
36 "stage5",
37 "stage6",
38 "stage7",
39 "stage8",
40 "stage9",
41 "stage10",
42 "stage11",
43 "stage12",
44 "stage13",
45 "stage14",
46 "stage15",
47 "stage16",
48 "stage17",
49 "stage18",
50 "stage19",
51 "stage20",
52 "stage21",
53 "stage22",
54 "stage23",
55 "stage24"
56 ],
57 "use_relative_position_bias": true
58 },
59 "backbone_hidden_size": 1024,
60 "batch_norm_eps": 1e-05,
61 "bin_centers_type": "softplus",
62 "bin_configurations": [
63 {
64 "max_depth": 10.0,
65 "min_depth": 0.001,
66 "n_bins": 64,
67 "name": "nyu"
68 },
69 {
70 "max_depth": 80.0,
71 "min_depth": 0.001,
72 "n_bins": 64,
73 "name": "kitti"
74 }
75 ],
76 "bin_embedding_dim": 128,
77 "bottleneck_features": 256,
78 "fusion_hidden_size": 256,
79 "head_in_index": -1,
80 "hidden_act": "gelu",
81 "initializer_range": 0.02,
82 "max_temp": 50.0,
83 "min_temp": 0.0212,
84 "model_type": "zoedepth",
85 "neck_hidden_sizes": [
86 256,
87 512,
88 1024,
89 1024
90 ],
91 "num_attractors": [
92 16,
93 8,
94 4,
95 1
96 ],
97 "num_patch_transformer_layers": 4,
98 "num_relative_features": 32,
99 "patch_transformer_hidden_size": 128,
100 "patch_transformer_intermediate_size": 1024,
101 "patch_transformer_num_attention_heads": 4,
102 "readout_type": "project",
103 "reassemble_factors": [
104 4,
105 2,
106 1,
107 0.5
108 ],
109 "torch_dtype": "float32",
110 "transformers_version": "4.42.0.dev0",
111 "use_batch_norm_in_fusion_residual": false,
112 "use_bias_in_fusion_residual": null,
113 "use_pretrained_backbone": false
114 }
115