About Training Efficiency.

[kuaiqushangzixiba]

"We evaluate the computational efficiency of LongSplat (Tab. 8), which achieves 281.71 FPS and trains in just 1 hour on an NVIDIA RTX 4090, nearly 30× faster than LocalRF. Our method also significantly reduces the model size to approximately 101 MB."

I noticed this paragraph, but it took me more than 10h to run the free dataset on 4090gpu. Does 1h here mean all the free data, or 1 scene out of 7, and 1 scene took me 2 hours, as shown in the log below.

grass log:
2025-08-27 11:03:05,784 - INFO: args: Namespace(eval=True, load_pose=False, sh_degree=3, feat_dim=32, n_offsets=10, voxel_size=0.1, update_depth=3, update_init_factor=16, update_hierachy_factor=4, init_frame_num=3, use_feat_bank=False, source_path='./data/free/grass', model_path='outputs/free/grass/baseline/2025-08-27_11:03:01', images='images', resolution=2, white_background=False, data_device='cuda', lod=0, appearance_dim=0, lowpoly=False, ds=1, ratio=1, undistorted=False, add_opacity_dist=False, add_cov_dist=False, add_color_dist=False, mode='free', iterations=30000, position_lr_init=0.0, position_lr_final=0.0, position_lr_delay_mult=0.01, position_lr_max_steps=30000, offset_lr_init=0.01, offset_lr_final=0.0001, offset_lr_delay_mult=0.01, offset_lr_max_steps=30000, feature_lr=0.0075, opacity_lr=0.02, scaling_lr=0.007, rotation_lr=0.002, mlp_opacity_lr_init=0.002, mlp_opacity_lr_final=2e-05, mlp_opacity_lr_delay_mult=0.01, mlp_opacity_lr_max_steps=30000, mlp_cov_lr_init=0.004, mlp_cov_lr_final=0.004, mlp_cov_lr_delay_mult=0.01, mlp_cov_lr_max_steps=30000, mlp_color_lr_init=0.008, mlp_color_lr_final=5e-05, mlp_color_lr_delay_mult=0.01, mlp_color_lr_max_steps=30000, mlp_featurebank_lr_init=0.01, mlp_featurebank_lr_final=1e-05, mlp_featurebank_lr_delay_mult=0.01, mlp_featurebank_lr_max_steps=30000, appearance_lr_init=0.05, appearance_lr_final=0.0005, appearance_lr_delay_mult=0.01, appearance_lr_max_steps=30000, percent_dense=0.01, lambda_dssim=0.2, loss_2d_correspondence_weight=1.0, depth_loss_weight=0.01, depth_l1_weight_init=0.1, depth_l1_weight_final=0.1, reproj_loss_weight_init=1.0, reproj_loss_weight_final=1.0, rotation_lr_init=0.001, translation_lr_init=0.001, start_stat=0, update_from=0, update_interval=100, update_until=0, pruning_ratio=0.0, min_opacity=0.005, success_threshold=0.8, densify_grad_threshold=0.0002, init_iteraion=3000, pose_iteration=200, local_iter=400, global_iter=900, window_size=5, post_iter=10000, convert_SHs_python=False, compute_cov3D_python=False, debug=False, ip='127.0.0.1', port=16357, debug_from=-1, detect_anomaly=False, test_iterations=[30000], quiet=False, gpu='-1')
2025-08-27 11:03:05,784 - INFO: Optimizing outputs/free/grass/baseline/2025-08-27_11:03:01
2025-08-27 11:57:12,604 - INFO:
[100 th] Evaluating train: L1 0.022611652864143253 PSNR 29.68408567428589
2025-08-27 12:51:25,914 - INFO:
[ITER 40000] Saving Gaussians
2025-08-27 12:51:28,632 - INFO:
Training complete.