Upload 2 files
Browse files- config.json +211 -0
- model.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,211 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"data": {
|
| 3 |
+
"augmentations": {
|
| 4 |
+
"blur_p": 0.2,
|
| 5 |
+
"flip_p": 0.5,
|
| 6 |
+
"gamma_p": 0.8,
|
| 7 |
+
"grayscale_p": 0.2,
|
| 8 |
+
"jitter_p": 0.8,
|
| 9 |
+
"random_blur": 2.0,
|
| 10 |
+
"random_gamma": 0.2,
|
| 11 |
+
"random_jitter": 0.4,
|
| 12 |
+
"random_scale": 2.0,
|
| 13 |
+
"shape_constraints": {
|
| 14 |
+
"height_min": 15,
|
| 15 |
+
"pixels_max": 600000,
|
| 16 |
+
"pixels_min": 200000,
|
| 17 |
+
"ratio_bounds": [
|
| 18 |
+
0.5,
|
| 19 |
+
2.5
|
| 20 |
+
],
|
| 21 |
+
"sample": true,
|
| 22 |
+
"shape_mult": 14,
|
| 23 |
+
"width_min": 15
|
| 24 |
+
},
|
| 25 |
+
"test_context": 1.0
|
| 26 |
+
},
|
| 27 |
+
"crop": "garg",
|
| 28 |
+
"data_root": "datasets",
|
| 29 |
+
"image_shape": [
|
| 30 |
+
480,
|
| 31 |
+
640
|
| 32 |
+
],
|
| 33 |
+
"normalization": "imagenet",
|
| 34 |
+
"num_copies": 2,
|
| 35 |
+
"num_frames": 1,
|
| 36 |
+
"sampling": {
|
| 37 |
+
"ETH3D": 1.0,
|
| 38 |
+
"Waymo": 1.0
|
| 39 |
+
},
|
| 40 |
+
"train_datasets": [
|
| 41 |
+
"ETH3D",
|
| 42 |
+
"Waymo"
|
| 43 |
+
],
|
| 44 |
+
"val_datasets": [
|
| 45 |
+
"IBims"
|
| 46 |
+
]
|
| 47 |
+
},
|
| 48 |
+
"eps": 1e-06,
|
| 49 |
+
"generic": {
|
| 50 |
+
"deterministic": true,
|
| 51 |
+
"seed": 13
|
| 52 |
+
},
|
| 53 |
+
"model": {
|
| 54 |
+
"expansion": 4,
|
| 55 |
+
"layer_scale": 1.0,
|
| 56 |
+
"name": "UniDepthV2",
|
| 57 |
+
"num_heads": 8,
|
| 58 |
+
"pixel_decoder": {
|
| 59 |
+
"depths": [
|
| 60 |
+
2,
|
| 61 |
+
2,
|
| 62 |
+
2
|
| 63 |
+
],
|
| 64 |
+
"dropout": 0.0,
|
| 65 |
+
"hidden_dim": 512,
|
| 66 |
+
"kernel_size": 3,
|
| 67 |
+
"name": "Decoder",
|
| 68 |
+
"out_dim": 64
|
| 69 |
+
},
|
| 70 |
+
"pixel_encoder": {
|
| 71 |
+
"cls_token_embed_dims": [
|
| 72 |
+
1024,
|
| 73 |
+
1024,
|
| 74 |
+
1024,
|
| 75 |
+
1024,
|
| 76 |
+
1024,
|
| 77 |
+
1024,
|
| 78 |
+
1024,
|
| 79 |
+
1024,
|
| 80 |
+
1024,
|
| 81 |
+
1024,
|
| 82 |
+
1024,
|
| 83 |
+
1024,
|
| 84 |
+
1024,
|
| 85 |
+
1024,
|
| 86 |
+
1024,
|
| 87 |
+
1024,
|
| 88 |
+
1024,
|
| 89 |
+
1024,
|
| 90 |
+
1024,
|
| 91 |
+
1024,
|
| 92 |
+
1024,
|
| 93 |
+
1024,
|
| 94 |
+
1024,
|
| 95 |
+
1024
|
| 96 |
+
],
|
| 97 |
+
"depths": [
|
| 98 |
+
6,
|
| 99 |
+
12,
|
| 100 |
+
18,
|
| 101 |
+
24
|
| 102 |
+
],
|
| 103 |
+
"embed_dim": 1024,
|
| 104 |
+
"embed_dims": [
|
| 105 |
+
1024,
|
| 106 |
+
1024,
|
| 107 |
+
1024,
|
| 108 |
+
1024,
|
| 109 |
+
1024,
|
| 110 |
+
1024,
|
| 111 |
+
1024,
|
| 112 |
+
1024,
|
| 113 |
+
1024,
|
| 114 |
+
1024,
|
| 115 |
+
1024,
|
| 116 |
+
1024,
|
| 117 |
+
1024,
|
| 118 |
+
1024,
|
| 119 |
+
1024,
|
| 120 |
+
1024,
|
| 121 |
+
1024,
|
| 122 |
+
1024,
|
| 123 |
+
1024,
|
| 124 |
+
1024,
|
| 125 |
+
1024,
|
| 126 |
+
1024,
|
| 127 |
+
1024,
|
| 128 |
+
1024
|
| 129 |
+
],
|
| 130 |
+
"freeze_norm": true,
|
| 131 |
+
"frozen_stages": 0,
|
| 132 |
+
"lr": 2e-06,
|
| 133 |
+
"name": "dinov2_vitl14",
|
| 134 |
+
"num_register_tokens": 0,
|
| 135 |
+
"output_idx": [
|
| 136 |
+
6,
|
| 137 |
+
12,
|
| 138 |
+
18,
|
| 139 |
+
24
|
| 140 |
+
],
|
| 141 |
+
"patch_size": 14,
|
| 142 |
+
"pretrained": null,
|
| 143 |
+
"stacking_fn": "last",
|
| 144 |
+
"use_norm": true,
|
| 145 |
+
"wd": 0.1
|
| 146 |
+
}
|
| 147 |
+
},
|
| 148 |
+
"training": {
|
| 149 |
+
"batch_size": 8,
|
| 150 |
+
"clipping": 1.0,
|
| 151 |
+
"cycle_beta": false,
|
| 152 |
+
"drop_path": 0.0,
|
| 153 |
+
"ema": true,
|
| 154 |
+
"f16": true,
|
| 155 |
+
"ld": 1.0,
|
| 156 |
+
"losses": {
|
| 157 |
+
"camera": {
|
| 158 |
+
"alpha": 1.0,
|
| 159 |
+
"fn": "l2",
|
| 160 |
+
"gamma": 1.0,
|
| 161 |
+
"input_fn": "linear",
|
| 162 |
+
"name": "Regression",
|
| 163 |
+
"output_fn": "sqrt",
|
| 164 |
+
"weight": 0.25
|
| 165 |
+
},
|
| 166 |
+
"confidence": {
|
| 167 |
+
"alpha": 1.0,
|
| 168 |
+
"fn": "l1",
|
| 169 |
+
"gamma": 1.0,
|
| 170 |
+
"input_fn": "linear",
|
| 171 |
+
"name": "Regression",
|
| 172 |
+
"output_fn": "sqrt",
|
| 173 |
+
"weight": 0.1
|
| 174 |
+
},
|
| 175 |
+
"depth": {
|
| 176 |
+
"dims": [
|
| 177 |
+
-2,
|
| 178 |
+
-1
|
| 179 |
+
],
|
| 180 |
+
"input_fn": "log",
|
| 181 |
+
"integrated": 0.15,
|
| 182 |
+
"name": "SILog",
|
| 183 |
+
"output_fn": "sqrt",
|
| 184 |
+
"weight": 1.0
|
| 185 |
+
},
|
| 186 |
+
"invariance": {
|
| 187 |
+
"name": "SelfDistill",
|
| 188 |
+
"output_fn": "sqrt",
|
| 189 |
+
"weight": 0.1
|
| 190 |
+
},
|
| 191 |
+
"ssi": {
|
| 192 |
+
"input_fn": "log1i",
|
| 193 |
+
"min_samples": 6,
|
| 194 |
+
"name": "EdgeGuidedLocalSSI",
|
| 195 |
+
"output_fn": "sqrt",
|
| 196 |
+
"use_global": true,
|
| 197 |
+
"weight": 1.0
|
| 198 |
+
}
|
| 199 |
+
},
|
| 200 |
+
"lr": 0.0001,
|
| 201 |
+
"lr_final": 1e-06,
|
| 202 |
+
"lr_warmup": 1.0,
|
| 203 |
+
"n_iters": 300000,
|
| 204 |
+
"nsteps_accumulation_gradient": 2,
|
| 205 |
+
"use_checkpoint": false,
|
| 206 |
+
"validation_interval": 2,
|
| 207 |
+
"warmup_iters": 75000,
|
| 208 |
+
"wd": 0.1,
|
| 209 |
+
"wd_final": 0.1
|
| 210 |
+
}
|
| 211 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ba73d3de735302ccc64a50f1e557122050c4b1893e6060b28dba05d6af3e67c6
|
| 3 |
+
size 1415383604
|