chunked/nltk_chunking/model_doc_vision-encoder-decoder/chunk_30.txt

As you can see, only 2 inputs are required for the model in order to compute a loss: pixel_values (which are the
images) and labels (which are the input_ids of the encoded target sequence).