chunked/nltk_chunking/model_doc_imagegpt/chunk_25.txt

Alternatively, one can further fine-tune the entire model on a downstream dataset, similar to BERT.