--- license: cc-by-nc-4.0 language: - zh --- * 使用lora对Qwen-14b-chat模型进行微调,使其能适应32k长度的上下文(目前仅支持中文)。 * 此项目仅提供lora参数,需要自行合并后使用。 ## 如何将lora参数合并到Qwen模型中 ```python from transformers import AutoModelForCausalLM, AutoTokenizer from transformers.generation import GenerationConfig from peft import PeftModel tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen-14B-Chat", trust_remote_code=True) model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen-14B-Chat", device_map="auto", trust_remote_code=True, bf16=True).eval() model=PeftModel.from_pretrained(model,"path_to_lora_weight") model=model.merge_and_unload() model.save_pretrained("save_path") ```