Do I Need to Set Up Softlink On Every Training with AMP?

I am reviewing the documentation on apex/examples/imagenet/. It states that a softlink should be set before the training execution.
(i.e
$ ln -sf /data/imagenet/train-jpeg/ train
$ ln -sf /data/imagenet/val-jpeg/ val)

I would like to deploy AMP in my own repo. Do I need to follow the example above to set up softlink with training data + validation data before I execute a training?

Also, I found that in main_amp.py, the argument "data" is a compulsory argument. (i.e. parser.add_argument('data', metavar='DIR', help='path to dataset')) Why in the example code, you don't need to specify the "data" argument? (e.g. $ python main_amp.py -a resnet50 --b 128 --workers 4 --opt-level O0 ./ )

Thank you.

Start by reading the documentation under apex/examples/imagenet/ and the argument definition in main_amp.py. Clarify whether the train and val softlinks are required for the example and explain how the example command supplies the compulsory data argument. Done means the documentation answers both questions consistently.