Open HitLuca opened 4 years ago
x2, would love a fix for this!!
Here you are @jfrench206, we just need to wait for a merge now
Afaik this usually isn't done because tensorflow<2
is split into two versions, tensorflow
and tensorflow_gpu
, and adding one as a requirement restricts users to one or the other.
I reckon that if the package by default uses tensorflow
and the user then installs tensorflow_gpu
on top of it, the project will use the gpu.
when installing this package, no tensorflow dependency is installed, because it's missing in the setup.py script. Given that it is a mandatory requirement, having it defined and locked to be a version prior to 2.0.0 would solve a bunch of issues