facebook/map-anything
9748k
1{2 "encoder_config": {3 "data_norm_type": "dinov2",4 "encoder_str": "dinov2",5 "gradient_checkpointing": false,6 "keep_first_n_layers": 24,7 "name": "dinov2_giant_24_layers",8 "norm_returned_features": false,9 "size": "giant",10 "torch_hub_force_reload": false,11 "uses_torch_hub": true,12 "with_registers": false13 },14 "geometric_input_config": {15 "cam_prob": 1.0,16 "cam_rot_encoder_config": {17 "enc_embed_dim": 1536,18 "encoder_str": "global_rep_encoder",19 "in_chans": 4,20 "name": "cam_rot_quats_encoder"21 },22 "cam_trans_encoder_config": {23 "enc_embed_dim": 1536,24 "encoder_str": "global_rep_encoder",25 "in_chans": 3,26 "name": "cam_trans_encoder"27 },28 "depth_encoder_config": {29 "apply_pe": false,30 "enc_embed_dim": 1536,31 "encoder_str": "dense_rep_encoder",32 "in_chans": 1,33 "name": "depth_encoder",34 "patch_size": 1435 },36 "depth_prob": 1.0,37 "depth_scale_norm_all_prob": 0.0,38 "dropout_prob": 0.0,39 "overall_prob": 1.0,40 "pose_scale_norm_all_prob": 0.0,41 "ray_dirs_encoder_config": {42 "apply_pe": false,43 "enc_embed_dim": 1536,44 "encoder_str": "dense_rep_encoder",45 "in_chans": 3,46 "name": "ray_dirs_encoder",47 "patch_size": 1448 },49 "ray_dirs_prob": 1.0,50 "scale_encoder_config": {51 "enc_embed_dim": 1536,52 "encoder_str": "global_rep_encoder",53 "in_chans": 1,54 "name": "scale_encoder"55 },56 "sparse_depth_prob": 0.0,57 "sparsification_removal_percent": 0.958 },59 "info_sharing_config": {60 "custom_positional_encoding": null,61 "model_return_type": "intermediate_features",62 "model_type": "alternating_attention",63 "module_args": {64 "custom_positional_encoding": null,65 "depth": 16,66 "dim": 1536,67 "distinguish_ref_and_non_ref_views": true,68 "gradient_checkpointing": false,69 "indices": [70 7,71 1172 ],73 "init_values": 1e-05,74 "input_embed_dim": 1536,75 "mlp_layer": "dummy",76 "mlp_ratio": 4,77 "name": "aat_16_layers_dinov2_vitg_init",78 "norm_intermediate": true,79 "num_heads": 24,80 "pretrained_checkpoint_path": null,81 "qk_norm": false,82 "qkv_bias": true,83 "size": "16_layers"84 }85 },86 "info_sharing_mlp_layer_str": "swiglufused",87 "load_specific_pretrained_submodules": false,88 "name": "mapanything",89 "pred_head_config": {90 "adaptor_config": {91 "dense_pred_init_dict": {92 "confidence_type": "exp",93 "confidence_vmax": Infinity,94 "confidence_vmin": 1,95 "depth_mode": "exp",96 "depth_vmax": Infinity,97 "depth_vmin": 0,98 "name": "raydirs+depth+pose+confidence+mask+scale",99 "ray_directions_clamp_min_of_z_dir": false,100 "ray_directions_mode": "linear",101 "ray_directions_normalize_to_unit_image_plane": false,102 "ray_directions_normalize_to_unit_sphere": true,103 "ray_directions_vmax": Infinity,104 "ray_directions_vmin": -Infinity,105 "ray_directions_z_dir_min": -Infinity106 },107 "input_dim": 6,108 "pose_pred_init_dict": {109 "cam_trans_mode": "linear",110 "cam_trans_vmax": Infinity,111 "cam_trans_vmin": -Infinity,112 "name": "raydirs+depth+pose+confidence+mask+scale",113 "quaternions_mode": "linear",114 "quaternions_normalize": true,115 "quaternions_vmax": Infinity,116 "quaternions_vmin": -Infinity117 },118 "scale_pred_init_dict": {119 "mode": "exp",120 "name": "raydirs+depth+pose+confidence+mask+scale",121 "vmax": Infinity,122 "vmin": 1e-08123 },124 "scene_rep_dim": 4,125 "scene_rep_type": "raydirs+depth+pose",126 "type": "raydirs+depth+pose+confidence+mask"127 },128 "adaptor_type": "raydirs+depth+pose+confidence+mask",129 "dpt_adaptor": {130 "confidence_type": "exp",131 "confidence_vmax": Infinity,132 "confidence_vmin": 1,133 "depth_mode": "exp",134 "depth_vmax": Infinity,135 "depth_vmin": 0,136 "name": "raydirs+depth+pose+confidence+mask+scale",137 "ray_directions_clamp_min_of_z_dir": false,138 "ray_directions_mode": "linear",139 "ray_directions_normalize_to_unit_image_plane": false,140 "ray_directions_normalize_to_unit_sphere": true,141 "ray_directions_vmax": Infinity,142 "ray_directions_vmin": -Infinity,143 "ray_directions_z_dir_min": -Infinity144 },145 "feature_head": {146 "checkpoint_gradient": false,147 "feature_dim": 256,148 "hooks": [149 0,150 1,151 2,152 3153 ],154 "input_feature_dims": [155 1536,156 1536,157 1536,158 1536159 ],160 "patch_size": 14161 },162 "gradient_checkpointing": false,163 "pose_adaptor": {164 "cam_trans_mode": "linear",165 "cam_trans_vmax": Infinity,166 "cam_trans_vmin": -Infinity,167 "name": "raydirs+depth+pose+confidence+mask+scale",168 "quaternions_mode": "linear",169 "quaternions_normalize": true,170 "quaternions_vmax": Infinity,171 "quaternions_vmin": -Infinity172 },173 "pose_head": {174 "input_feature_dim": 1536,175 "num_resconv_block": 2,176 "patch_size": 14,177 "rot_representation_dim": 4178 },179 "regressor_head": {180 "checkpoint_gradient": false,181 "input_feature_dim": 256,182 "output_dim": 6183 },184 "scale_adaptor": {185 "mode": "exp",186 "name": "raydirs+depth+pose+confidence+mask+scale",187 "vmax": Infinity,188 "vmin": 1e-08189 },190 "scale_head": {191 "input_feature_dim": 1536,192 "output_dim": 1193 },194 "type": "dpt+pose"195 },196 "pretrained_checkpoint_path": null,197 "specific_pretrained_submodules": [],198 "torch_hub_force_reload": false,199 "use_register_tokens_from_encoder": true200}