With all due respect which part of your model is this going to be added to, the training or the testing part?

[Taonanana]

noise_schedule = NoiseScheduleVP(schedule='linear', continuous_beta_0=0.1, continuous_beta_1=20.)
            model_fn = GaussianDiffusion()
            # model_fn = model_wrapper(
            #     diffusion_model,
            #     noise_schedule,
            #     model_type="noise",  # or "x_start" or "v" or "score"
            #     model_kwargs={},
            # )
            dpm_solver = DPM_Solver(model_fn, noise_schedule, algorithm_type="dpmsolver++")
            x_sample = dpm_solver.sample(
                x_T, //what is this?
                t_end=1e-4,
                steps=20,
                order=3,
                skip_type="time_uniform",
                method="multistep",
            )

[LuChengTHU]

Hi @Taonanana ,

Sorry for the late reply. I don't exactly understand the question you're asking. Could you please give me a more detailed explanation?