On the NeRF run being killed while executing run_custom.py #206
Description
Dear author, thank you for open-sourcing your work and for your contributions.
When I run on my own dataset, I keep encountering
[bundlesdf.py] First nerf run, create Runner, latest nerf frame frame_0293
Killed.
A portion of the log is shown below:
`frame_0000 frame_0018 frame_0027 frame_0036 frame_0044
[2025-12-08 09:35:12.713] [warning] [Bundler.cpp:920] OptimizerGPU begin, global_corres#=3192
global_corres=3192
maxNumResiduals / maxNumberOfImages = 51192 / 5 = 10238
m_maxNumberOfImages*m_maxCorrPerImage = 5 x 1461 = 7305
m_solver->solve Time difference = 7.888[ms]
[2025-12-08 09:35:12.740] [warning] [Bundler.cpp:924] OptimizerGPU finish
[2025-12-08 09:35:12.740] [warning] [Bundler.cpp:320] Added frame frame_0044 as keyframe, current #keyframe: 5
[bundlesdf.py] processNewFrame done frame_0044
[bundlesdf.py] frame_0044 prepare data for nerf
[bundlesdf.py] out_dir: /data/2025-12-6_singleframe//frame_0044/nerf
[tool.py] compute_scene_bounds_worker start
[tool.py] compute_scene_bounds_worker done
[tool.py] merge pcd
[tool.py] compute_translation_scales done
translation_cvcam=[-0.0057411 0.00371899 -0.00244663], sc_factor=3.501653900906107
[bundlesdf.py] First nerf run, create Runner, latest nerf frame frame_0044
Killed
(py38) root@docker-desktop:/BundleSDF# [nerf_runner.py] Octree voxel dilate_radius:1
level 0, resolution: 16
level 1, resolution: 32
level 2, resolution: 64
level 3, resolution: 128
GridEncoder: input_dim=3 n_levels=4 level_dim=2 resolution=16 -> 128 per_level_scale=2.0000 params=(2462192, 2) gridtype=hash align_corners=False
sc_factor 2.4511577306342747
translation [-0.0057411 0.00371899 -0.00244663]
[nerf_runner.py] denoise cloud
[nerf_runner.py] Denoising rays based on octree cloud
[nerf_runner.py] bad_mask#=2
rays torch.Size([64093, 12])
[bundlesdf.py] Start training, latest nerf frame frame_0044
[nerf_runner.py] train progress 0/501
[nerf_runner.py] Iter: 0, valid_samples: 393216/393216, valid_rays: 2048/2048, loss: 4.9221530, rgb_loss: 3.8475151, rgb0_loss: 0.0000000, fs_rgb_loss: 0.0000000, depth_loss: 0.0000000, depth_loss0: 0.0000000, fs_loss: 0.5076177, point_cloud_loss: 0.0000000, point_cloud_normal_loss: 0.0000000, sdf_loss: 0.5670201, eikonal_loss: 0.0000000, variation_loss: 0.0000000, truncation(meter): 0.0100000, pose_reg: 0.0000000, reg_features: 0.0000000,
[nerf_runner.py] train progress 50/501
[nerf_runner.py] train progress 100/501
[nerf_runner.py] train progress 150/501
[nerf_runner.py] train progress 200/501
[nerf_runner.py] train progress 250/501
[nerf_runner.py] train progress 300/501
[nerf_runner.py] train progress 350/501
[nerf_runner.py] train progress 400/501
[nerf_runner.py] train progress 450/501
[nerf_runner.py] train progress 500/501
[bundlesdf.py] Training done, latest nerf frame frame_0044
[bundlesdf.py] Getting mesh
[nerf_runner.py] query_pts:torch.Size([4330747, 3]), valid:69748
[nerf_runner.py] Running Marching Cubes
[nerf_runner.py] done V:(10980, 3), F:(21952, 3)`
I’m really curious why the Killed error keeps happening — my GPU is an RTX 4060 Laptop.
What’s even stranger is what happens to the program logs after it’s killed, and I also can’t find any corresponding output files in the output folder.