Issue with finetuning a seq-to-seq model

[brando]

does this solve this? How to add all standard special tokens to my tokenizer and model?