Which SKImage version is compatible with Numpy 1.19.5

Hi, I work with the release 1.19.5 of Numpy and would like to know which release of scikit-image is compatible with this version.
For several reasons I cannot upgrade to latest release of Numpy and when I install the latest release of scikit-image I have an error on ndarray size.
So I am wondering if using a previous release of scikit-image could solve the problem and if so which version do you suggest ?
Thanks for your help

Following is the message I receive by just importing skimage :
ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

Hey @watchever, thanks for reaching out and sorry for the slow response.

Our most recent version scikit-image, right now still v0.19.3, should work with NumPy 1.19.5 (see requirements/default.txt for tag 0.19.3).

Maybe your error stems from elsewhere. How did you install scikit-image and NumPy?

That error happens if you use an sklearn that is compiled against a NumPy that is newer then the one in use.

So whatever you did to install sklearn, it may be you just have to redo it or so to make sure it uses the old NumPy. SciPy wheels should already be compiled that way, so what @lagru said is important: How did you install both?

1 Like

Thanks guys for your help and answers. At least I reinstall last version of numpy and skimage and no more error message. Not clear to me what’s happened, but well… it works. Thanks again