but yeah, I am training classifiers in both cases, the same memory-saving techniques could potentially be applied to the recent transformers for tabular data.

→ View original post on X — @rasbt