Getting a eError: Default process group has not been initialized, please make sure to call init_process_group

22

Why would you have to train the model again if you save baseline’s state_dict? You would be able to create the baseline model afterwards and load its state_dict, wouldn’t you?