chinese-llama-7b-ift is trained from Llama-7b with Chinese and English Corpus and an ift dataset. The tokenizer is from ziqingyang/chinese-llama-lora-7b