Can't compile PyTorch 0.4.1 with CUDA support on Linux; commands not found

3

Do you by any change used a windows computer to get the code before putting it on the machine where you compile?

Nope.

Will you need the distributed package? If not, try compiling with NO_DISTRIBUTED=1 as MPI might be the problem here.

Yea, I don’t think I’ll be trying to distribute training anytime soon. I set this variable and it compiled and installed fine this time. Tanks a lot for help!