I’m getting failed builds during CI with the error No module named 'torch'. For my settings.ini file, I followed the fast.ai repo and made different pip/Conda requirements for installing Pytorch.
You need to actually modify your setup.py file to install those bits for you, by default it only looks in requirements. You can look at fastai’s setup.py file for inspriation
Sorry, I have been having the same problem. None of the conda_requirements within my settings.ini file are being installed by GitHub Cl. I have had a look at fastai’s setup.py but it does not mention conda anywhere. Could you possibly provide any further suggestions as to how you resolved your issue? Thanks for your help!