Conda or PIP install tensorflow-gpu on Windows in 2023?

General Discussion
,
#1

I upgraded my Windows PC to a 13th gen Intel CPU and Nvidia 40-series GPU, and found myself have to install tensorflow-gpu all over again. In the past, I used conda install tensorflow-gpu V2.6 and it worked fine. I also tried WSL2 before and found Windows is faster. So I’d like to stick to Windows.

I tried to use conda to install tensorflow-gpu this time around. The installation was fine. But as soon as I import tensorflow as tf, I receive the following errors.

So what is the right way to install tensorflow-gpu on Windows in July 2023?

Errors are shown below:
AttributeError: module ‘numpy’ has no attribute ‘object’.
np.object was a deprecated alias for the builtin object. To avoid this error in existing code, use object by itself. Doing this will not modify any behavior and is safe.

#2

Hi @kdlin1, Could you please try to install the numpy version 1.23.4 using pip install numpy==1.23.4 in your environment and let us know if the issue gets resolved or not. Thank You.

#4

Thanks, @Kiran_Sai_Ramineni I did a conda install of numpy=1.23.4. The problem went away! I really appreciate the pointer.

I am now still at tensorflow-gpu=2.6.0. What features am I missing with this version?