When evaluating trained T5-base on the generated test set of `SQuAD`

[zhaochenyang20]

Token indices sequence length is longer than the specified maximum sequence length for this model (973 > 512). Running this sequence through the model will result in indexing errors