[BUG] UnpicklingError while fitting model

[jbottke]

Describe the bug
When trying to fit a Mambular classification model on the adult census dataset, the training process results in an UnpicklingError.

To Reproduce

• load adult census dataset using shapiq with shapiq.datasets.load_adult_census()
• preprocess data using sklearns train_test_split() function which shuffles dataset und splits it into four numpy arrays (x_train, x_test, y_train, y_test)
• create Instance of MambularClassifier and fit it on numpy arrays x_train, y_train

Expected behavior
Method is expected to train a Mambular classification model and save the best model under model_checkpoints as a .ckpt file.

Screenshots

Desktop:

• OS: Ubuntu via WSL2
• Python version 3.10.12
• Mambular Version 1.1.0

[AnFreTh]

Thanks for raising this issue! Could you please add your PyTorch and lightning versions?

[jbottke]

• lightning 2.5.0.post0
• lightning-utilities 0.12.0
• pytorch-lightning 2.5.0.post0
• torch 2.6.0
• torchmetrics 1.6.1

[AnFreTh]

@mkumar73 Restricting torch to 2.2.2 - 2.5.1 should solve the issue.

[AnFreTh]

Thanks again for raising this issue! The latest release v1.2.0 #222 should solve this issue. Feel free to reopen or open any other issues if you encounter any more bugs.