variational-diffusion-models icon indicating copy to clipboard operation
variational-diffusion-models copied to clipboard

PyTorch implementation of Variational Diffusion Models.

Results 2 variational-diffusion-models issues
Sort by recently updated
recently updated
newest added

according to doi.org/10.48550/arXiv.2401.06281, 4.2, equation 4.62, image prediction loss can be parametrized by adding the snr_t term to [pred_loss ](https://github.com/addtt/variational-diffusion-models/blob/90a8b59d16845f8fb52e4587d149073caf3465fb/vdm.py#L110C9-L110C18). Thus for image prediction in [VDM.forward()](https://github.com/addtt/variational-diffusion-models/blob/90a8b59d16845f8fb52e4587d149073caf3465fb/vdm.py#L110C9-L110C62) we change: ``` -...