| config: {'model': {'encoder': {'name': 'dcsplat', 'input_image_shape': [518, 518], 'head_mode': 'pcd', 'same_head': False, 'num_level': 3, 'gs_param_dim': 256, 'camera_prompt': False, 'num_visual_prompt': 0, 'use_extrinsic_token': False, 'use_intrinsic_token': False, 'extrinsic_drop': 0.1, 'intrinsic_drop': 0.1, 'use_pnp_solver': False, 'align_corners': False, 'image_scale': 1.0, 'use_uncertainty_token': True, 'gs_bf16': True, 'aggregator_bf16': False, 'img_num_frame_token_N': 0, 'img_num_global_token_N': 0, 'use_gs_param_img_num_embed': True, 'use_mean_param_img_num_embed': False}, 'decoder': {'name': 'splatting_cuda', 'background_color': [0.0, 0.0, 0.0], 'make_scale_invariant': False, 'use_gsplat': True}, 'density_control': {'name': 'density_control_module', 'mean_dim': 32, 'gs_param_dim': 256, 'refinement_layer_num': 1, 'num_level': 3, 'grad_mode': 'absgrad', 'use_mean_features': True, 'refinement_type': 'quadtree', 'refinement_hidden_dim': 32, 'aggregation_mode': 'mean', 'num_heads': 1, 'score_mode': 3, 'sigmoid_score': True, 'latent_dim': 128, 'num_latents': 64, 'num_self_attn_per_block': 2}}, 'render_loss': {'mse': {'weight': 1.0}, 'lpips': {'weight': 0.05, 'apply_after_step': 0}}, 'density_control_loss': {'complex_score': {'weight': 0.01, 'log_scale': False, 'grad_scale': 5000.0, 'mode': 1}}, 'direct_loss': {'l1': {'weight': 0.8}, 'ssim': {'weight': 0.2}}, 'wandb': {'project': 'DCSplat', 'entity': 'scene-representation-group', 'name': '1027_DL3DV_FULL_U', 'mode': 'online', 'tags': ['dl3dv', '224x448', 'full']}, 'mode': 'train', 'data_loader': {'train': {'num_workers': 16, 'persistent_workers': True, 'batch_size': 4, 'seed': 1234}, 'test': {'num_workers': 4, 'persistent_workers': False, 'batch_size': 1, 'seed': 2345}, 'val': {'num_workers': 1, 'persistent_workers': True, 'batch_size': 1, 'seed': 3456}}, 'optimizer': {'lr': 0.0002, 'warm_up_steps': 125, 'backbone_lr_multiplier': 0.1, 'backbone_trainable': 'GB+H', 'accumulate': 1}, 'checkpointing': {'load': None, 'every_n_train_steps': 9375, 'save_top_k': 1, 'save_weights_only': True}, 'train': {'extended_visualization': False, 'print_log_every_n_steps': 10, 'camera_loss': 1.0, 'one_sample_validation': 'datasets/dl3dv/room_batch.pt', 'align_corners': False, 'intrinsic_scaling': True, 'train_context': True, 'train_hybrid': False, 'use_pred_point_focal': False, 'scaling_detach': True, 'complex_threshold_range': [0.0, 3.0], 'use_umeyama': False, 'pred_focal_cam_loss': False, 'direct_loss_weight': 0.0, 'direct_iteration_steps': 0, 'twice_backward': False, 'use_uncertainty_camera_weight': True, 'use_target_uncertainty': 'point_similarity_mult', 'verbose': False, 'complex_random_mode': 'num_gs_mode3', 'start_density_control': 0.0, 'base_complex_exp_rate': 4.0, 'use_refine_aux': False, 'train_extrinsic': True}, 'test': {'output_path': 'test/full/dl3dv', 'align_pose': False, 'pose_align_steps': 100, 'rot_opt_lr': 0.005, 'trans_opt_lr': 0.005, 'compute_scores': True, 'save_image': False, 'save_video': False, 'save_active_mask_image': False, 'save_complex_score_image': False, 'save_compare': False, 'pred_intrinsic': True, 'complex_threshold': 0.4, 'complex_threshold_list': [0.0, 0.2, 0.4, 0.6, 0.8], 'threshold_mode': 'ratio', 'nvs_view_N_list': [3, 6, 16, 32, 64]}, 'seed': 111123, 'trainer': {'max_steps': 18751, 'val_check_interval': 500, 'gradient_clip_val': 0.5, 'num_nodes': 1}, 'dataset': {'dl3dv': {'make_baseline_1': False, 'relative_pose': True, 'augment': True, 'background_color': [0.0, 0.0, 0.0], 'overfit_to_scene': None, 'skip_bad_shape': True, 'view_sampler': {'name': 'bounded_dl3dv', 'min_target_views': 4, 'num_context_views': 16, 'warm_up_steps': 0, 'max_img_per_gpu': 16, 'min_distance_to_context_views': 0, 'initial_min_distance_between_context_views': 5, 'initial_max_distance_between_context_views': 7, 'min_distance_between_context_views': 32, 'max_distance_between_context_views': 256, 'min_gap_multiplier': 3, 'max_gap_multiplier': 5, 'initial_min_gap_multiplier': 2, 'initial_max_gap_multiplier': 3}, 'name': 'dl3dv', 'roots': ['datasets/dl3dv'], 'input_image_shape': [224, 448], 'original_image_shape': [270, 480], 'cameras_are_circular': False, 'baseline_min': 0.001, 'baseline_max': 100.0, 'max_fov': 100.0, 'avg_pose': False, 'rescale_to_1cube': False, 'intr_augment': False, 'normalize_by_pts3d': False}}, '_wandb': {}} |
