TypeError: expected np.ndarray (got numpy.ndarray)

[AyushSomani001]

The code from the Notebook (notebooks/image_predictor_example.ipynb) works fine in Google Collab:

But, when I'm trying to run in the server (NVIDIA A100-SXM4-40GB; CUDA Version: 12.5), there's a image processing issue.

I did check the version and passing implicitly with numpy.array(), especially numpy versions are the same. I'm not sure why the issue persist!

[ronghanghu]

Hi @AyushSomani001, I think this is likely because you have a numpy installation or a PIL installation that is not compatible with your local PyTorch installation.

Could you try the following and see if it works?

import numpy as np
import torch

torch.from_numpy(np.zeros((3, 1024, 1024), dtype=np.float32))

It should be resolved by reinstalling numpy and pytorch with versions that match each other

[seandkim]

It should be resolved by reinstalling numpy and pytorch with versions that match each other

How do I find out which numpy version matches with a specific pytorch version? I am currently using pytorch 2.5.1

[seandkim]

@AyushSomani001 I was able to get it working by downgrading numpy=1.26.4

I am still curious what is the best way to find the matching version, I was only able to find it through this stackoverflow post