Hi, thank you for releasing the test code and model. I'm playing with the "inference_single_video" script, finding that using diffusion steps > 1000 introduces noticeable Gaussian noise. And from another post, I saw that timestep_respacing is the flag to change the denoising steps. How shall I tune them?
Hi, thank you for releasing the test code and model. I'm playing with the "inference_single_video" script, finding that using diffusion steps > 1000 introduces noticeable Gaussian noise. And from another post, I saw that timestep_respacing is the flag to change the denoising steps. How shall I tune them?