tensorflow
Fresh install doesn't work (incompatible versions of tensorflow and tensorflow_privacy)
Problem: Fresh install of tensorflow_privacy doesn't work
Steps to recreate: Create a new environment, run
pip install tensorflow-privacy
Now from within python, run
import tensorflow_privacy
I get this horrendous-looking error
ValueError: Arg specs do not match: original=FullArgSpec(args=['input', 'dtype', 'name', 'layout'], varargs=None, varkw=None, defaults=(None, None, None), kwonlyargs=[], kwonlydefaults=None, annotations={}), new=FullArgSpec(args=['input', 'dtype', 'name'], varargs=None, varkw=None, defaults=(None, None), kwonlyargs=[], kwonlydefaults=None, annotations={}), fn=<function ones_like_v2 at 0x16fe7a200>
Quick fix:
pip install tfp-nightly
Root of the problem: This issue in tfp shows that the problem is that the current version of tfp was built using tf 2.13, however the latest version of tf is now 2.14 and they are incompatible. This is fixed in the nightly version of tfp but not yet in the main release.
SO, the error is not your fault (the privacy team) but in terms of the user experience I installed privacy and it didn't work. Is this something you could fix by specifying version numbers? I think if you do nothing the problem will fix itself when a new version of tfp is released, but then it might reappear when a new version of tensorflow is released
