chunked/nltk_chunking/model_doc_owlvit/chunk_17.txt

Open-vocabulary classification is enabled by replacing the fixed classification layer weights with the class-name embeddings obtained from the text model.