Python 3 使用fire实现命令行传参

在完成源码编写后，有些类或者函数需要通过命令行传递参数，可以使用fire工具包：
下面的代码中，如果仅需要给add传递参数，将add暴露给命令行的方法只需要在文件末尾添加：fire.Fire()
对类中的parse函数，需要暴露给命令的方法为在文件末尾添加：fire.Fire(DefaultConfig)
一般不同时暴露上面的两个。

import fire
class DefaultConfig(object):
    def __init__(self):
        self.env = 'default' # visdom environment
        self.model = 'AlexNet' # model name must be the same with models in 'models/__init__.py'

        self.train_data_root = '/home/sjtuer/Desktop/DRL/S6/data/train'
        self.test_data_root = '/home/sjtuer/Desktop/DRL/S6/data/test'
        self.load_model_path = None # 'checkpoint/model.py' # load the pre-trained model, if None: no load

        self.batch_size = 128
        self.use_gpu = True
        self.num_workers = 4 # how many workers for loading data
        self.print_freq = 20 # print info every N batch

        self.debug_file = '/tmp/debug' # if os.path.exists(debug_file): enter ipdb
        self.result_file = 'result.csv'

        self.max_epoch = 10
        self.lr = 0.001
        self.lr_decay = 0.95 # when val_loss increase, lr = lr * lr_decay
    weight_decay = 1e-4 # loss function

    def parse(self, **kwargs):
        '''
        update configuration based on kwargs dictionary
        '''
        for k,v in kwargs.items():
            if not hasattr(self,k):
                warning.warn("Warnnin: opt has not attribut %s" %k)
            setattr(self, k, v)
        # print(self.__dict__)
        # print(self.__class__)
        print('user config:')
        for k, v in self.__dict__.items(): # __class__ 表示当前操作的对象的类
            # print(k)
            if not k.startswith('__'):
                print(k, getattr(self, k))
opt = DefaultConfig()
def add(a,b):
    return a+b


if __name__ == '__main__':
    dc = DefaultConfig()
    print(dc.batch_size)
    dc.parse()
    print(dc.batch_size)
    fire.Fire()