RAG for FEVER Dataset

I had a few queries on running rag for FEVER?

  1. In the finetuning step as i understand from the paper for fever they would first regenerate the claim. So for seq2seq format would our “train.source” and “train.target” would be just the same claim?

  2. And then how do we actually make the model classify/give a fever label?

  3. In addition how do we after fine tuning, run just the code in inference mode where can either generate answers for Question Asnwering or a label for fever?

Any help would be much appreciated