Why? Their models are useful as a starting place for doing your own fine tuning, but they have never stood well on their own.
they always undertrain the model
they always undertrain the model