CUDA 12.1, Python 3.11, Pytorch 2.5.1, Torchvision 0.20.1

[atonalfreerider]

I just wanted to confirm that it is possible to successfully run this amazing program using more up-to-date libraries. This is the environment.yml that I used to successfully run the tests:

name: tracks_to_4d
channels:
  - pytorch
  - nvidia
  - conda-forge
  - defaults
dependencies:
  - pip
  - python=3.11
  - pytorch=2.5.1
  - pytorch-cuda=12.1
  - torchvision=0.20.1
  - pip:
    - tqdm
    - opencv-python
    - matplotlib
    - einops
    - imageio
    - imageio[ffmpeg]