Using Python argparse: How to handle command line arguments

1. Preface

argparse is a command line parameter parsing package for python. It is convenient to use when the code needs to frequently modify parameters. The main usage is Enter the parameters you want to modify on the command line.

2. Common frameworks for argparse usage

import argparse

def get_parser():

    # argparse.ArgumentParser生成argparse对象 description为描述信息，当在命令行输入需要显示帮助信息时，会显示

    parser = argparse.ArgumentParser(description="pytorch unet training")

    # 路径参数设置 help为参数的帮助信息

    parser.add_argument("--data_path", default="./", help="DRIVE root")

    # 预测类别数量 type如果不指定需要输入的是str类型

    parser.add_argument("--num_classes", default=1, type=int)

    # 指定设备使用

    parser.add_argument("--device", default="cuda", help="training device")

    # 指定batch size大小 "-b", "--batch_size"表示两个都可以在命令行使用

    parser.add_argument("-b", "--batch_size", default=4, type=int)

    return parser

if __name__ =='__main__':

    parser = get_parser()

    args = parser.parse_args()

    print(args)

As mentioned above, get_parser() is some of the usage methods introduced in this article. This python file is called python_argparse_test1.py, in which

parser = argparse.ArgumentParser(description="pytorch unet training") is used to create a parser object

add_argument() is used to add parameters

args = parse_args() in parser.parse_args() to obtain parsing Parameters

1. Get the parameter list

When python python_argparse_test1.py is entered on the command line, print the parameters obtained by args and get:

Namespace(batch_size=4, data_path ='./', device='cuda', num_classes=1)

Indicates the parameter list parsed by this parameter parser

2. Get help information

Input python python_argparse_test1.py -h or python python_argparse_test1.py --help displays information, where usage shows its usage, pytorch unet training is the description when creating the object, and below is the information and usage of each parameter

3. Command line modification parameters

import argparse

def get_parser():

    # argparse.ArgumentParser生成argparse对象 description为描述信息，当在命令行输入需要显示帮助信息时，会显示

    parser = argparse.ArgumentParser(description="pytorch unet training")

    # 路径参数设置 help为参数的帮助信息 default为默认参数

    parser.add_argument("--data_path", default="./", help="DRIVE root")

    # 预测类别数量 type如果不指定需要输入的是str类型

    parser.add_argument("--num_classes", default=1, type=int)

    # 指定设备使用

    parser.add_argument("--device", default="cuda", help="training device")

    # 指定batch size大小 "-b", "--batch_size"表示两个都可以在命令行使用

    parser.add_argument("-b", "--batch_size", default=4, type=int)

    return parser

if __name__ =='__main__':

    parser = get_parser()

    args = parser.parse_args()

    print("data_path: ",args.data_path)

    print("num_classes: ", args.num_classes)

Command line input: python python_argparse_test1.py --data_path Desktop --num_classer 4, the result is as follows:

data_path: Desktop

num_classes: 4

You can see that the parameters have been modified through the command line

4. The use of '_' and "__"

if __name__ =='__main__':

    parser = get_parser()

    args = parser.parse_args()

    print("-b: ",args.b)

    print("--batch_size: ", args.batch_size)

Enter python python_argparse_test1.py -b 10 --batch_size 20 on the command line, and an error will be reported:

This is because when '_' and '__' exist at the same time, The system defaults to the parameter name

. Change the above code to:

if __name__ =='__main__':

    parser = get_parser()

    args = parser.parse_args()

    print("--batch_size: ",args.batch_size)

but the command line will not be affected. Continue to execute the command python python_argparse_test1.py -b 10 to get:

--batch_size: 10

5. Type usage

type will force the input command line characters to be converted into type type

if __name__ =='__main__':

    parser = get_parser()

    args = parser.parse_args()

    print("--batch_size type: ",type(args.batch_size))

Command line input: python python_argparse_test1 .py --batch_size '10' gets:

--batch_size type:

6.required: used to indicate whether this parameter needs to be provided

parser.add_argument("--num_classes", default=1, type=int, required=True)

If you enter the command python python_argparse_test1.py

This will report an error prompting the parameters that are indeed required

python_argparse_test1.py: error: the following arguments are required: --num_classes

7.choicesSelect parameters

parser.add_argument('-arch', required=True, choices=['alexnet', 'vgg'])

If you run the command: python python_argparse_test1.py -arch cnn

This will report an error

python_argparse_test1.py: error: argument - arch: invalid choice: 'cnn' (choose from 'alexnet', 'vgg')

The above is the detailed content of Using Python argparse: How to handle command line arguments. For more information, please follow other related articles on the PHP Chinese website!