Multifit / Fast.ai 2.0

Hello Community,

Are there any plans to integrate Multifit for Fast.ai 2.0?

The model architecture has been quite beneficial for languages like German, where QRNN and subwords tokenization works better than the AWD LSTM. As I understand now, as Fast AI 2.0 is a complete rewrite, I can only use V1 with Multifit.

Thanks in advance for any replies!
Have a nice friday!

1 Like