I'd drop pandas if polars worked eamlessly with sklearn.

Most of the time it works, but there are still corner cases throughout the python ML ecosystem where polars fails.

This and the decision not to natively plot with matplotlib keep me in pandas for most tasks. (Plus there's is still a relatively large demand for pandas training.)

You can do to_pandas() and from_pandas() which doesn't have much overhead, or to_numpy() directly which is typically zero-copy and thus no overhead: https://docs.pola.rs/api/python/stable/reference/dataframe/a...