Tensorflow 1.14.0 and numpy 1.16.0 compatibility with gensim

Question

I've already downloaded

1. tensorflow 1.14.0
2. numpy 1.16.0

when using pip install gensim, the gensim vision is 4.1.2 and it needs numpy >=1.17.0

Then my problem is which one the gensim vision i should install to match tensorflow 1.14.0 and numpy 1.16.0.

Answer 1

Can you clarify why it's important to you to use an old numpy (1.16.0 is from over 3 years ago), and not even a version with priority fixes (like 1.16.6, released in December 2019)?

Generally, it's best to use the latest version of major open-source projects, for fixes, performance, & support from the wider community – unless there's a very specific known reason that your purposes require an older version.

(Similarly, TensorFlow seems to be at release 2.8.0.)

Further, after you've picked a few of the key libraries for your exact needs, it's usually best to let your environment-installation software (pip install or similar) pick the right prerequisite versions of other libraries. For example, pip install gensim will typically install not just the latest version of Gensim, but also whatever specific numpy it needs – if not already present.

And, doing all this in a project-specific 'environment' (as via the Python3 venv facility, or Conda environments) can help avoid confusion about what's needed for your system's Python installation, and what your specific project needs to use.

What code will you be running, or writing, and what libraries do you expect it needs, in what versions? I suspect if you start a fresh environment, & begin by installing just the exact libraries you know you'll need, the right other prerequisite libraries will arrive. And if they don't, and you get specific error message you can share here, it'll be easier to recommend improvements.)

Answer 2

Now gensim should be lunched under the Numpy>=1.17 ver. So I suggest you upgrade the Numpy