AttributeError: 'super' object has no attribute 'sklearn_tags' (sklearn code version mismatch)

[yuyiyi]

I got the following error when running the hippocampus demo and the step-by-step tutorial in google colab.

cebra_posdir3 = cebra_posdir3_model.transform(hippocampus_pos.neural)

error
AttributeError Traceback (most recent call last)
in <cell line: 2>()
1 cebra_posdir3_model.fit(hippocampus_pos.neural, hippocampus_pos.continuous_index.numpy())
----> 2 cebra_posdir3 = cebra_posdir3_model.transform(hippocampus_pos.neural)

4 frames
/usr/local/lib/python3.10/dist-packages/sklearn/base.py in sklearn_tags(self)
857
858 def sklearn_tags(self):
--> 859 tags = super().sklearn_tags()
860 tags.transformer_tags = TransformerTags()
861 return tags

AttributeError: 'super' object has no attribute 'sklearn_tags'

Is there a solution for this?

[gbradham]

My workaround is using this line of code
!pip install --upgrade git+https://github.com/AdaptiveMotorControlLab/CEBRA.git@main
I believe this was fixed like a month ago. Also if you are using the preloaded datasets don't just replace the original install with this one because I believe there is an error when trying to download the .jl files.

[yuyiyi]

Thanks @gbradham !
upgrade the cebra package did not solve this issue.
Instead, I installed an earlier version of sklearn.

[gbradham]

Nice. What was your pipeline of pip installs? I believe I had gotten it to work as well by downgrading the sklearn version as well. And I wonder why that did not work, I had trouble for a while before installing from the git repo. Best of luck!

[yuyiyi]

@gbradham I'm working on a windows 11 PC with cuda. After pip install cebra in the anaconda environment, I did pip install scikit-learn==1.5.2

[gbradham]

Nice. I did have problems with windows and anaconda so if you face similar issues I found that just removing the cebra package from the conda env and then reinstalling would work. Can I also ask which demo notebook you were working on?

[MMathisLab]

Hi all! thanks for this discussion! We will release a new pip version of CEBRA this week, which will fix any of these issues due to other package updates, cheers!