Renxi Wang

Padding for training and inference

5

Is llama 2 trained with batch? If so, why there is no pad token? If I want to finetune the model and then inference in batch. A suggestion is to...