RuntimeError when running custom ScanNet dataset

[Planetinaline]

Hi and thanks for your excellent work ! I was testing a custom dataset created from scannet(simply using the 'color' folder in ScanNet scenes as the 'input' folder in custom dataset) and encountered a RuntimeError which I couldn't figure out, so i came here seeking help. Here is the error i received when trying to run python train.py -s /home/ubuntu/newdisk/cwk/PGSR/data/scene0003_00 -m output --max_abs_split_points 0 --opacity_cull_threshold 0.05: Optimizing output Output folder: output [08/09 20:06:31] Tensorboard not available: not logging progress [08/09 20:06:31] Reading camera 2/2 [08/09 20:06:31] Converting point3d.bin to .ply, will happen only the first time you open the scene. [08/09 20:06:31] xyz (0, 3) [08/09 20:06:31] cameras_extent 5.5 [08/09 20:06:31] Loading Training Cameras [08/09 20:06:31] load camera 1Loading Test Cameras [08/09 20:06:34] computing nearest_id [08/09 20:06:34] Number of points at initialisation : 0 [08/09 20:06:34] Traceback (most recent call last): File "train.py", line 495, in training(lp.extract(args), op.extract(args), pp.extract(args), args.test_iterations, args.save_iterations, args.checkpoint_iterations, args.start_checkpoint, args.debug_from) File "train.py", line 97, in training scene = Scene(dataset, gaussians) File "/home/ubuntu/newdisk/cwk/PGSR/scene/init.py", line 128, in init self.gaussians.create_from_pcd(scene_info.point_cloud, self.cameras_extent) File "/home/ubuntu/newdisk/cwk/PGSR/scene/gaussian_model.py", line 181, in create_from_pcd dist = torch.sqrt(torch.clamp_min(distCUDA2(torch.from_numpy(np.asarray(pcd.points)).float().cuda()), 0.0000001)) RuntimeError: tabulate: failed to synchronize: cudaErrorInvalidConfiguration: invalid configuration argument

[Planetinaline]

seems that it's a problem with colmap, it would crash when doing sparse reconstruction resulting in wrong points3D.ply