Could you share some details on what algorithm you use to generate the puzzles?
The original T5 models were trained with masked language modelling similar to BERT and afterwards also trained on a autoregressive task similar to GPT.
I would be very suprised if this hasn't already been implemented since GPT-3/3.5
Could you share some details on what algorithm you use to generate the puzzles?
The original T5 models were trained with masked language modelling similar to BERT and afterwards also trained on a autoregressive task similar to GPT.
I would be very suprised if this hasn't already been implemented since GPT-3/3.5