kaihe

got nan loss trying to use pytorch-lightning mix precision API ``` trainer_kwargs["precision"] = 16 trainer_kwargs["amp_backend"] = "native" ```

@joanrod I give it up mix precision training

@GioFic95 refer to [this line](https://github.com/CompVis/latent-diffusion/blob/main/models/ldm/bsr_sr/config.yaml#L56). second_stage_model of SR ddpm has no encode function, therefore cond_stage_key image is still in **Image Space** not **Latent Space**. Hence this line ` elif self.conditioning_key...

> @GioFic95 refer to [this line](https://github.com/CompVis/latent-diffusion/blob/main/models/ldm/bsr_sr/config.yaml#L56). second_stage_model of SR ddpm has no encode function, therefore cond_stage_key image is still in **Image Space** not **Latent Space**. Hence this line ` elif...

After a few QR scan, I finally managed to log in : )