This issue comes due to incompatibility of numpy with some other pre-installed packages in your virtual environment. Your numpy would have been upgraded to latest version by one of your user package installations (this other package, say tf, would have recently been upgraded and when you install it's latest version, it's pushing numpy to latest version as well. with this being done, packages, apart from numpy, remain same and would be compatible with older version of numpy).

• You should downgrade the package (her tf or anyother package that you are installing) by specifying an older version, as the problem is not with numpy.
• if you create new virtual environment, you don't need to do anything about numpy.
• However, if you want to rectify old environment then, you have to manually downgrade numpy as well(with tf).

NOTE: I have not mentioned versions here because this is a recurring problem with new releases, so you downgrade to a version that is older (say year or year and a half).