* Allow config cahnge * Allow val change in wandb config * Don't resume transfer learning runs * Add entity in log dataset
torch.distributed.run
gloo
check_file()
normalize=True