Initial weight ensemble training?

[ShihengDuan]

Hi all,

I'm using PyTorch-lightning to rewrite some of my training scripts. I wonder if there is any function to automate the initial-weight ensemble training process. For now, I'm using randomly initialized weight so I have to train 10 times and get the ensemble predictions. Ideally, these 10 models could be trained in parallel. Is there any function in lightning that can automate this process?

Thanks!