Basic use of argparse Library

argparse Kuo is python The next command line parameter management Library , Support int、str、float、bool、 Array etc. 5 Basic data types . Default values are also supported when parsing command line parameters . Besides , It also supports modifying command line parameters , Add command line parameters and other operations in the code . adopt argparse library , It can realize simple management of parameters , Control some process or parameter values in code execution .

Basic usage

1、int Support

argparse Writing ,class_nums The default value is 7

parser.add_argument("--class_nums", type=int, default=7)

Command line parameter passing （ Suppose the name of the program file is test.py）, A new value was passed , Override default

test.py --class_nums 8

2、str Support

argparse Writing ,load_from No default value is set , If you want to set it, please set default

parser.add_argument("--load_from", type=str, help='load trained model to generate predictions of validation set')

Command line parameter passing （ Suppose the name of the program file is test.py）, A new value was passed "best.pth" Assign a value to load_from

test.py --load_from "best.pth"

3、float Support

argparse Writing ,lr The default value is =0.0003

parser.add_argument("--lr", type=float, default=0.0003)

Command line parameter passing （ Suppose the name of the program file is test.py）, A new value was passed , Override default

test.py --lr 0.001

4、bool Support

argparse Writing ,pretrained The default value is None.store_true Indicates that the incoming parameter is True

parser.add_argument("--pretrained", action="store_true")

Command line parameter passing （ Suppose the name of the program file is test.py）, here pretrained The value of is True

test.py --pretrained

5、 Array support

argparse Writing ,input_shape The default value is [256, 256].nargs Indicates that there may be multiple values for the incoming parameter , That is to say, the parameters passed in are not necessarily 2 individual ,3 One can also

parser.add_argument('--input_shape', nargs='+', type=int, default=[256, 256])

Command line parameter passing （ Suppose the name of the program file is test.py）, here input_shape The value of is [512, 512]

test.py --input_shape 512 512

Full call

The following code realizes the parameter value , Modify parameter value , Add a complete call to the parameter value . The default parameters cannot be modified by executing the following code in a non console environment , Parameters can only be modified by executing under the console

import argparse
class Options:
    def __init__(self):
        parser = argparse.ArgumentParser(' Namespace ')
        parser.add_argument("--class_nums", type=int, default=7)
        parser.add_argument("--pretrained", action="store_true")
        parser.add_argument("--lr", type=float, default=0.0003)
        parser.add_argument("--load_from", type=str,
                            help='load trained model to generate predictions of validation set')
        parser.add_argument('--input_shape', nargs='+', type=int, default=[256, 256])
        self.parser = parser

    def parse(self):
        args = self.parser.parse_args()
        return args
if __name__ == "__main__":
    args = Options().parse()
    print("\n Original parameters ：")
    print(args)
    
    print("\n Take the parameter value ：")
    print("args.class_nums:",args.class_nums)
    
    print("\n Modify the parameters ：")
    args.input_shape=[512,512]
    print(args)
    
    print("\n The new parameters ：")
    args.new_key=[512,512]
    print(args)

Blogger test code ：python .\test.py --input_shape 128 128
The execution output is as follows ：

Original parameters ：
Namespace(class_nums=7, input_shape=[128, 128], load_from=None, lr=0.0003, pretrained=False)
Take the parameter value ：
args.class_nums: 7
Modify the parameters ：
Namespace(class_nums=7, input_shape=[512, 512], load_from=None, lr=0.0003, pretrained=False)
The new parameters ：
Namespace(class_nums=7, input_shape=[512, 512], load_from=None, lr=0.0003, new_key=[512, 512], pretrained=False)