I had a slight clarification. In each round, you are adding the new data to the old, saving as a json file and training from scratch (with model initialized to resnet checkpoints), right?
As in the training command for each round, the resume-checkpoint and the auto-resume have not been passed as arguments.
Thank you very much for your code.
I had a slight clarification. In each round, you are adding the new data to the old, saving as a json file and training from scratch (with model initialized to resnet checkpoints), right?
As in the training command for each round, the resume-checkpoint and the auto-resume have not been passed as arguments.