Open hhelm10 opened 4 years ago
After looking over the entire package, I found out that proglearn
itself really doesn't use tensorflow
or scipy
at all. We only use them for benchmarks and tutorials, so I believe they probably should not count as package dependencies.
@jdey4 @levinwil How do you guys think?
it may be good to have "lighter" package requirements.. i.e. do not have keras or any model-specific requirement in requirement.txt in case users only want to use forests (say) and do not want to install any of the deep learning packages.