Installation steps on a brand new PC:
- install python / pip
- install VS community edition
- install latest cuda software (https://developer.download.nvidia.com/compute/cuda/12.4.1/local_installers/cuda_12.4.1_551.78_windows.exe)
- install pytorch (pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121)
- pip install fastbook
Expected:
everything functions normally
Actual:
fastbook needs an older version of pytorch, so it uninstalls pytorch, and re-installs the NON-CUDA version
Can you somehow detect if the wrong version of pytorch is installed to see whether it’s CUDA or not, rather than installing the CPU only version?