Chinese version of GPT2 training code, using BERT tokenizer. - View it on GitHub