SOURCE RECORD · github
Chinese version of GPT2 training code, using BERT tokenizer.
Open canonical source ↗
Research paper status → · Book status →