The interpretability of using the well-trained Kan 2.0 model for inference on individual samples.

[Franklanqing]

Hello everyone, I am trying to understand the interpretability of using the well-trained Kan 2.0 model for inference on individual samples. I attempted to use model.attribute for interpretability explanations of a single sample, but I found that while the model loads successfully and can predict the correct category, the output of model.attribute is all nan. Can anyone help me with this issue? Thank you.