The default torch installs GPU support and the total packages size is somewhere around 5GB just for torch and depedencies. If GPU support is not required, then using the cpu (non gpu) library saves a lot of space and install time. The requirements.txt (or pip install command) can be update with an alternative URL to point to the CPU wheel
# Use CPU Torch for a smaller install
--extra-index-url https://download.pytorch.org/whl/cpu
torch
We might also include an environment variable or other option when installing repast4py to switch the behavior in the repast4py requirements.txt
The default torch installs GPU support and the total packages size is somewhere around 5GB just for torch and depedencies. If GPU support is not required, then using the cpu (non gpu) library saves a lot of space and install time. The requirements.txt (or pip install command) can be update with an alternative URL to point to the CPU wheel
We might also include an environment variable or other option when installing repast4py to switch the behavior in the repast4py requirements.txt