Closed beniz closed 8 months ago
This PR allows choosing between L1 and MSE for multiscale loss in diffusion model training.
This PR allows choosing between L1 and MSE for multiscale loss in diffusion model training.