RuntimeError: module compiled against API version 0xe but this version of numpy is 0xd

[RafaD5]

Hello,

I installed the latest version of MLJAR (0.10.2). When I import the package the following error arrises

RuntimeError: module compiled against API version 0xe but this version of numpy is 0xd

I created a new enviroment for the installation with conda. The steps I followed for the installation were:

conda create -n mljar39 python=3.9
conda activate mljar39
pip install mljar-supervised

Then I run from supervised.automl import AutoML and got the RuntimeError

Then, as suggested in this issue, I updated numpy (explanation here) but got the following error

And now, when I import MLJAR the error does not longer arise and I can use MLJAR, but I don't know if this solution is reliable.

[pplonski]

@RafaD5 I believe that there is a small mismatch between numpy versions for some packages. I've added Optuna to MLJAR and it doesn't support the newest numpy version.

The package will work even with the warning. I observed the warning on Ubuntu 16.04 but for Ubuntu 20.04 it works without warning. Hard to point exact fix. However, the numpy update should help.

I would recommend waiting for an update on the Optuna package.

[xuzhang5788]

I used the same method to create a conda env. I update mljar-supervided with pip install -U mljar-supervised. When I ran from supervised.automl import AutoML, I still got
RuntimeError: module compiled against API version 0xe but this version of numpy is 0xd
My Ubuntu is 20.04

[pplonski]

I've updated Optuna (2.7.0) and Numpy (>=1.20.0) packages in requirements.txt. The warning should disappear.

Changes will go to 0.10.4. To use current changes please install package from GitHub:

pip install -q -U git+https://github.com/mljar/mljar-supervised.git@master

[CyFeng16]

If you are getting above error in python3.8 environment, python3.7 works fine.