google/t5-large-ssm-nqo

The model was pre-trained using T5's denoising objective on C4 . Pre-trained on REALM's salient span and then fine-tuned on Natural Questions (NQ) The model can be used as follows for closed book question answering: t5_tok = AutoTokenizer.from_pretrained("google/t5-large-ssm-nqo") T5_qa_model = AutoModelForSeq2SeqLM, AutoTokenized = t5-tok . The model is used to answer questions without access to any external context or knowledge . To facilitate reproducibility and future work, we release our code and,