大模型微调报错二

训练大模型Qwen15-05B-Chat-GPTQ-Int4
训练使用qwen1.5 sft：
命令：python finetune.py --model_name_or_path /llm/Qwen15-05B-Chat-GPTQ-Int4
–output_dir ./checkpoints
–model_max_length 512
–data_path /data/agi/dataset/train_0.5M_CN/output600.jsonl
–use_lora True
–per_device_train_batch_size 1
–q_lora True
–learning_rate 5e-4
运行报错：
ValueError: Found modules on cpu/disk. Using Exllama backend requires all the modules to be on GPU.You can deactivate exllama backend by setting disable_exllama=True in the quantization config object
处理：
1） 修改finetune.py。

    model = AutoModelForCausalLM.from_pretrained(
        model_args.model_name_or_path,
        config=config,
        cache_dir=training_args.cache_dir,
        device_map=device_map,
        quantization_config=GPTQConfig(
            bits=4,
	    disable_exllama=True)  # 添加修改