Missing argument for --pretrained_model in training.py

[inesvanlaeys]

Hi,

I've noticed that in the training.py script, there seems to be a missing argument for --pretrained_model in the argument parser. The code currently contains:

parser.add_argument(
    "",
    default='',
    type=str,
    help="Path of pretrained model. If empty, the model will be trained from scratch.",
)

However, the argument name is empty. I believe this should be:

parser.add_argument(
    "--pretrained_model",
    default='',
    type=str,
    help="Path of pretrained model. If empty, the model will be trained from scratch.",
)

Could you please confirm if this is the intended behavior and if the correction is needed?

Thanks for your help!