nick kartashov

same issue

same problem

exactly

I think first approach with train and predict methods more standard and clean, like in sklearn. Without that feature many ml techniques like stacking,blending become not so trivial. About performance...

Sorry, I'm not saw your todo . Thank you very much for future work.

It was solved when I commented out these lines: !pip uninstall flash-attn !pip install --no-build-isolation flash-attn==2.3.0