Could you provide the script from ImageNet pre-training? I have looked into the checkpoint. However, the arguments in the checkpoint are not very clear to me. For instance:
There are several drop probabilities - 'drop=0.2, drop_block=None, drop_connect=0.2, drop_path=0.1'. How are them integrated in the training?
There are some arguments - 'ratio=[0.75, 1.3333333333333333], recount=1', 'lr_cycle_limit=1, lr_cycle_mul=1.0', 'reprob=0.2', 'scale=[0.08, 1.0]'. What are the meanings of them?
Could you provide the script from ImageNet pre-training? I have looked into the checkpoint. However, the arguments in the checkpoint are not very clear to me. For instance:
Many thanks for considering my request.