Summarization failed when using t5

[Amayama]

Dear ohmeow, thanks for your convenient framework. It helps me a lot.
However, a problem occurs when I try to apply the t5-base model to summarization task using seq2seq settings in the training procedure of https://ohmeow.github.io/blurr/text-modeling-seq2seq-summarization.html. I changed the Tokenizer and config to
T5TokenizerFast | T5ForConditionalGeneration and the model was trained smoothly. However, when I tried to evaluate after first epoch, an exception occured as TypeError: Exception occured in ``Seq2SeqMetricsCallback`` when calling event ``after_batch``: forward() got an unexpected keyword argument 'prefix'. I wonder why this exception happened and still cannot find solution although I checked my code and core clearly. It would be a great pleasure if U could give me some helps.

[ohmeow]

Could you put your code in a gist that I could look at?