Use random port for initialize PyTorch for training?

[huangchieh]

RuntimeError: The server socket has failed to listen on any local network address. The server socket has failed to bind to [::]:12357 (errno: 98 - Address already in use). The server socket has failed to bind to 0.0.0.0:12357 (errno: 98 - Address already in use).

When I try to use multiple submit script, this error occurs. The solution might be using a random port.

[huangchieh]

In the beginning of run, there are two lines of code to configure the port.

https://github.com/SINGROUP/StyleTransAugment/blob/5c667e8bcacd544126b91fdfa784bce68e86e22e/4_fit_posnet_augrate.py#L201C1-L203C40

    # Initialize the distributed environment. 
    os.environ['MASTER_ADDR'] = 'localhost'
    os.environ['MASTER_PORT'] = '12357'

I change to this

StyleTransAugment/4_fit_posnet_augrate.py

Line 203 in 9495b7d

os.environ['MASTER_PORT'] = '{}'.format(random.randint(10000, 20000) # Random port

    os.environ['MASTER_ADDR'] = 'localhost'
    os.environ['MASTER_PORT'] = '{}'.format(random.randint(10000, 20000) # Random port