Views
No views yet
1from transformers import AutoModelForCausalLM, AutoTokenizer
2from transformers.generation import GenerationConfig
3from peft import PeftModel
4
5tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen-14B-Chat", trust_remote_code=True)
6model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen-14B-Chat", device_map="auto", trust_remote_code=True, bf16=True).eval()
7
8model=PeftModel.from_pretrained(model,"path_to_lora_weight")
9model=model.merge_and_unload()
10model.save_pretrained("save_path")