Configuration parameters

model_args,data_args,train_args,finetuning_args,generating_args = get_train_args()

Load participle tokenizer

tokenizer = AutoTokenizer.from_pretrained(model_path)

Read the config file

config = AutoConfig.from_pretrained(model_path)

Reading model files

model = AutoModelForCausalLM.from_pretrainded(model_path,config)

Initialize the adapter to the model

model = init_adapter(model,args)

model = model.train()

Train

trianer = SftTrainer(model,args,tokenizer)

##Merge the weights of the original model with the weights of the adapter to generate your own model. from transformers import AutoTokenizer,AutoModelForCausalLM form peft import PeftModel import torch

model_name_or_path = "your_LLM_model_path" adapter_name_or_path = "your_lora_model_path" save_path = "save_model_path"

tokenizer = AutoTokenizer.from_pretrained( model_name_or_path, trust_remote_code=True ) model = AutoModelForCausalLM.from_pretrained( model_name_or_path, trust_remote_code=True, low_cpu_mem_usage=True, torch_dtype=torch.float16, device_map="auto" ) model = PeftModel.from_pretrained(model, adapter_name_or_path) model = model.merge_and_unload()

tokenizer.save_pretrained(save_path) model.save_pretrained(save_path)

##If you have more information about Lora and LLM fine-tuning problem,please contact me at [email protected]