chunked/nltk_chunking/model_doc_perceiver/chunk_29.txt

As the Perceiver's input
length will not have an impact on the computation time of the self-attention layers, one can provide raw bytes,
providing inputs of length 2048 to the model.