Image-to-3D
English
wala
mult-iview-to-3d
kamalrahimi commited on
Commit
4e82918
1 Parent(s): 67342c9

Create args.json

Browse files
Files changed (1) hide show
  1. args.json +182 -0
args.json ADDED
@@ -0,0 +1,182 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "strategy": null,
3
+ "gpu_workers": 8,
4
+ "restore_path": null,
5
+ "prefetch_factor": null,
6
+ "matual_precision": "medium",
7
+ "max_concurrency": 1500,
8
+ "multipart_size": 8388608,
9
+ "print_every": 100,
10
+ "save_every": 100,
11
+ "validation_every": 15000,
12
+ "validation_every_log": 500,
13
+ "visualization_every": 15000,
14
+ "log_level": "info",
15
+ "experiment_type": "max",
16
+ "experiment_every": 5,
17
+ "fast_dev_run": false,
18
+ "limit_val_batches": 0.05,
19
+ "ft_train_number": null,
20
+ "ft_train_datasets": null,
21
+ "val_cnt": null,
22
+ "use_even_val": false,
23
+ "use_ema": true,
24
+ "ema_decay": 0.9999,
25
+ "use_ema_weights": true,
26
+ "use_compile": true,
27
+ "batch_size": 8,
28
+ "test_batch_size": 8,
29
+ "num_workers": 23,
30
+ "test_threshold": 0.05,
31
+ "num_gpus": 8,
32
+ "seed": 1,
33
+ "epochs": 300,
34
+ "optimizer": "Adam",
35
+ "lr": 0.0001,
36
+ "train_mode": "train",
37
+ "use_local_storage": false,
38
+ "auto_exp_name": null,
39
+ "input_type": "Wavelet",
40
+ "output_type": "Wavelet",
41
+ "encoder_type": "General_Encoder_Down_2",
42
+ "decoder_type": "General_Decoder_Up_2",
43
+ "encoder_num_tran": 0,
44
+ "decoder_num_tran": 0,
45
+ "last_feature_transform": null,
46
+ "reconstruct_loss_type": "mean",
47
+ "quantizer_type": "original",
48
+ "normalize_latent": null,
49
+ "e_dim": 4,
50
+ "n_e": 1024,
51
+ "beta": 0.25,
52
+ "sample_mode": "bilinear",
53
+ "padding": 0.1,
54
+ "gamma": 1,
55
+ "grid_size": 12,
56
+ "t_loss": 1.0,
57
+ "num_latent_tokens": 256,
58
+ "dataset_name": "all",
59
+ "voxel_transform": null,
60
+ "num_points": 2048,
61
+ "num_sdf_points": 5000,
62
+ "categories": null,
63
+ "resolution": 256,
64
+ "max_depth": 3,
65
+ "max_training_level": 2,
66
+ "point_num": 16384,
67
+ "keep_level": 2,
68
+ "data_keep_level": 2,
69
+ "wavelet": "bior6.8",
70
+ "padding_mode": "constant",
71
+ "use_normalization": false,
72
+ "use_shift_mean": false,
73
+ "start_stage": 0,
74
+ "use_adaptive_stage_update": false,
75
+ "no_rebalance_loss": true,
76
+ "use_compact_indices": true,
77
+ "sample_threshold_ratio": 0.03125,
78
+ "use_batched_threshold": true,
79
+ "use_sample_training": true,
80
+ "use_sample_threshold": true,
81
+ "div_hyp": 1.0,
82
+ "checkpoint": null,
83
+ "use_timestamp": false,
84
+ "num_iterations": 300000,
85
+ "gpu": "0",
86
+ "threshold": 0.45,
87
+ "sampling_type": null,
88
+ "auto_precision": "bf16",
89
+ "gradient_clip_val": 1.0,
90
+ "dropout": 0.0,
91
+ "sdf_points": 20000,
92
+ "sdf_sample_type": "mixture",
93
+ "sdf_res": 256,
94
+ "greater_or_no": true,
95
+ "s3_bucket": "build3d-wavelets",
96
+ "s3_prefix": "dataset",
97
+ "use_s3": true,
98
+ "wavelet_transform": "all",
99
+ "test_file_name": "model_performance.csv",
100
+ "network_type": "latent_uvit",
101
+ "diffusion_beta_schedule": "cosine",
102
+ "diffusion_step": 1000,
103
+ "diffusion_rescale_timestep": 100,
104
+ "diffusion_scale_ratio": 1.0,
105
+ "diffusion_model_var_type": "FIXED_SMALL",
106
+ "diffusion_model_mean_type": "START_X",
107
+ "diffusion_loss_type": "MSE",
108
+ "diffusion_sampler": "second-order",
109
+ "dit_block_type": "cross_dit",
110
+ "att_patch_size": 1,
111
+ "att_hidden_size": 1152,
112
+ "transformer_num_blocks": 32,
113
+ "transformer_num_heads": 16,
114
+ "transformer_add_num_register": 0,
115
+ "unet_model_channels": 128,
116
+ "unet_num_res_blocks": [
117
+ 3
118
+ ],
119
+ "learnable_skip_r": null,
120
+ "add_condition_res_ch": 128,
121
+ "with_fix_pos": true,
122
+ "cond_num_mapping_layers": 0,
123
+ "add_condition_time_ch": true,
124
+ "add_condition_input_ch": null,
125
+ "use_pointcloud_conditions": false,
126
+ "pc_encoder_type": "PointNet_Simple",
127
+ "use_pointvoxel_encoder": false,
128
+ "num_pc_points": 2500,
129
+ "use_pc_samples": false,
130
+ "sample_num": 2500,
131
+ "pc_dims": 1024,
132
+ "num_inds": 1024,
133
+ "pc_output_dim": 1024,
134
+ "use_voxel_conditions": false,
135
+ "voxel_context_dim": 1024,
136
+ "voxel_dim": 8,
137
+ "voxel_resolution": 16,
138
+ "use_image_conditions": true,
139
+ "use_camera_index": true,
140
+ "render_resolution": 384,
141
+ "max_images_num": 55,
142
+ "image_transform": "dino",
143
+ "clip_model_type": "dino-l-14_reg",
144
+ "input_view_cnt": 4,
145
+ "use_multiple_views_grids": true,
146
+ "training_views": [
147
+ 3,
148
+ 6,
149
+ 10,
150
+ 26
151
+ ],
152
+ "testing_views": null,
153
+ "use_depth_conditions": false,
154
+ "use_wavelet_conditions": false,
155
+ "dp_cond": 0.05,
156
+ "scale": 3,
157
+ "guidance_type": null,
158
+ "dp_cond_type": null,
159
+ "use_autoencoder": false,
160
+ "checkpoint_type": "last_ft_10000",
161
+ "pre_quant": true,
162
+ "latent_normalization": false,
163
+ "use_autoencoder_ema": false,
164
+ "filter_path": "filter_list",
165
+ "finetune": false,
166
+ "finetune_dp_cond": null,
167
+ "finetune_dp_cond_type": null,
168
+ "test_log_num": null,
169
+ "precision": "bf16",
170
+ "gradient_clip_val_2": 1.0,
171
+ "wavelet_transform_2": null,
172
+ "weight_decay": 0.0,
173
+ "opt_eps": 1e-08,
174
+ "latent_lr": 0.0001,
175
+ "pin_memory": false,
176
+ "n_px": 224,
177
+ "cond_emb_dim": 1024,
178
+ "cond_grid_size": 256,
179
+ "cond_grid_emb_size": 1024,
180
+ "condition_dim": 1024,
181
+ "num_cond_vectors": 256
182
+ }