checkpoint creation for finetuned stable diffusion model I am trying to save the finetuned parameter checkpoints after each epoch for a stable diffusion model. If I use this: model_state_dict': pipe.unet.state_dict(), #saves the entire model. While this model_state_dict': fine_tuned_state_dict will not save nothing 64bytes, optimizer values also look wrong. Not sure if the problem is that the optimizer is not updating any of the parameters or my checkpoint saving is not correct. The loss value also seems to fluctuate and not steadily go down. I...