I just copied the slim folder in my object detection folder and then renamed all the modules so that there is no **from object-detection import blabla ** . BUT that is obviously the dumbest way to do it, you may supply something on the lines of
pwd/slim to your command in floyd then python train.py and so fort.
something like that might work , i am not a linux guy ,but nonetheless :
floyd run -- gpu blablabla "export PYTHONPATH=$PYTHONPATH:
pwd/slim && python train.py blablba .. &&
cp training/* /output/ " if you want to save your trained model .