fudan-zvg / 4d-gaussian-splatting

[ICLR 2024] Real-time Photorealistic Dynamic Scene Representation and Rendering with 4D Gaussian Splatting
MIT License
605 stars 44 forks source link

sh_degree_t error when running train.py #26

Open 1943055307 opened 8 months ago

1943055307 commented 8 months ago

Hello. There are several questions i want to ask.

  1. Is there any change in module "diff-gaussian-rasterization"? I failed to install the module "diff-gaussian-rasteriazation" with Ninja, so i use another virtual environment with original 3dgaussian and "diff-gaussian-rasterization". I am not sure whether it works which depends on whether there is any change in module "diff-gaussian-rasterization"
  2. In the way I described in que1, I can run train.py but is says raceback (most recent call last): File "E:\Paper\Dynamic3DGaussian\Real_Time_Photorealistic_Dynamic_Scene_Representation_and_Rendering_with_4D_Gaussian_Splatting\4d-gaussian-splatting\train.py", line 402, in <module> training(lp.extract(args), op.extract(args), pp.extract(args), args.test_iterations, args.save_iterations, args.start_checkpoint, args.debug_from, File "E:\Paper\Dynamic3DGaussian\Real_Time_Photorealistic_Dynamic_Scene_Representation_and_Rendering_with_4D_Gaussian_Splatting\4d-gaussian-splatting\train.py", line 108, in training render_pkg = render(viewpoint_cam, gaussians, pipe, background) File "E:\Paper\Dynamic3DGaussian\Real_Time_Photorealistic_Dynamic_Scene_Representation_and_Rendering_with_4D_Gaussian_Splatting\4d-gaussian-splatting\gaussian_renderer\__init__.py", line 37, in render raster_settings = GaussianRasterizationSettings( TypeError: GaussianRasterizationSettings.__new__() got an unexpected keyword argument 'sh_degree_t'. I check the code mentioned in the error message but found everything correct. So I don't know what the error is. I guess that may be related to the problem I mentioned in que1.
  3. Is there any other user who succeed run the program following the steps in the README.md with the given environment.yml. Could you please provide the information about your CUDA toolkit version and torch version? Sincere thanks.
1943055307 commented 8 months ago

As for question 1 and 3. I use cudatoolkit11.7 and torch 1.13.0 with python 3.10. And there is always error when i use ninja to construct the module "diff-gaussian-rasterization". So i am considering uninstall 11.7 and try 11.6 which is used by the author. But I'm not sure whether will it work...