chunked/nltk_chunking/model_doc_vision-encoder-decoder/chunk_23.txt

To perform inference, one uses the [generate] method, which allows to autoregressively generate text.