Llama推理报参数校验错误TypeError: The input value must be int. but got 'NoneType.

经验分享 Tech Blogs 模型训练-Model Training
1

1. 系统环境

硬件环境(Ascend/GPU/CPU): Ascend
MindSpore版本: 不限
执行模式（PyNative/ Graph）：不限

2. 问题描述

Llama推理报参数校验错误

(MindSpore) [ma-user transfer_checkpoint]$python test_baichuan_78.py  
Traceback (most recent call last):  
    File "/home/ma-user/work/mindformers/transfer_checkpoint/test_baichuan_78.py", 1ine: 16, in <module>  
        baichuan_model - LlamaForCausalLM(  
    File "/home/ma-user/anaconda3/envs/MindSpore/1ib/python3.9/site-packages/mindformers/models/1lama/1lama .py", 1ine 253, in __init__  
        self.model - LlamaModel (config=config)  
    File "/home/ma-user/anaconda3/envs/MindSpore/1ib/python3.9/site-packages/mindformers/models/1lama/1lama.py", line 94, in __init__  
        Validator.check_positive_int(config.batch_size)  
    File "/home/ma-user/anaconda3/envs/MindSpore/1ib/python3.9/site-packages/mindspore/_checkparam•py", line 331, in check_positive_int  
        return _check number(arg value, 0, GT,.int, arg-name, prim name)  
    File -/home/ma-user/anaconda3/envs/Mindspore/1ib/python3.9/site-packages/mindspore/ checkparam.py", line 220, in check number  
        check param()  
    File -/home/ma-user/anaconda3/ervs/MindSpore/1ib/python3.9/site-packages/mindspore/ checkparam.py",line 208, in check param  
        raise TypeError( f"{prim info} must be farg type. __name__}, but got {type(arg_value).__name}'")  
TypeError: The input value must be int. but got 'NoneType.

3. 解决方案

出错位置在做参数校验，上面是对config.batch_size做校验，检验结果为config.batch_size为int，但是实际是NoneType，表示传入的值是一个空值；需要检查config.batch_size是否配置，在yaml中配置位置是否错误，值是否正确；
其他参数校验错误，也可参照以上解决办法。