Loss for training KL-VAE

LTH14 / mar

PyTorch implementation of MAR+DiffLoss https://arxiv.org/abs/2406.11838

MIT License

766 stars 40 forks source link

Closed Vinnieassaulter closed 1 month ago

Vinnieassaulter commented 1 month ago

Hello! May I ask if you used VQ-16 similar perceptual loss or just reconstruction loss and KL divergence when training KL-VAE?

LTH14 commented 1 month ago

Hi! Please see the configuration here https://github.com/CompVis/latent-diffusion/blob/main/models/first_stage_models/kl-f16/config.yaml#L7-L12. Just the same as LDM's KL-VAE, we use reconstruction loss, perceptual loss, KL divergence, and GAN loss.