Hi, I saw the scores for t5-base and bart models using implicit knowledge get around 19-20 rougeL, and was having issues matching that score. Currently, I am getting around 11 rougeL. I was wondering what hyperparameters and scheduler were used? Thanks
Hi, I saw the scores for t5-base and bart models using implicit knowledge get around 19-20 rougeL, and was having issues matching that score. Currently, I am getting around 11 rougeL. I was wondering what hyperparameters and scheduler were used? Thanks