1 2 3 4
_target_: text_recognizer.models.transformer.TransformerLitModel interval: null monitor: val/loss ignore_tokens: ["<s>", "<e>", "<p>"]