Is it safe to use pytorch-cpu as the backend on a machine without a gpu?

If i run conda install - c fastai- c pytorch fastai it installs the GPU version of fastai on the machine. Granted, it still works even without a GPU but I’m wondering if i can get away with using pytorch-cpu instead on that machine since it’ll save space and maybe have less overhead.

I’ll probably have to have conda ignore dependencies or install fastai via pip if i do it this way.