1 2 3 4 5 6 7
_target_: text_recognizer.models.transformer.TransformerLitModel interval: step monitor: val/loss ignore_tokens: [ <s>, <e>, <p> ] start_token: <s> end_token: <e> pad_token: <p>