sampler_tonemap_rescalecfg.py

import torch


class TonemapNoiseWithRescaleCFG:
    @classmethod
    def INPUT_TYPES(s):
        return {"required": {"model": ("MODEL",),
                             "tonemap_multiplier": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 100.0, "step": 0.01}),
                             "rescale_multiplier": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01}),
                             }}
    RETURN_TYPES = ("MODEL",)
    FUNCTION = "patch"

    CATEGORY = "custom_node_experiments"

    def patch(self, model, tonemap_multiplier, rescale_multiplier):

        def tonemap_noise_rescale_cfg(args):
            cond = args["cond"]
            uncond = args["uncond"]
            cond_scale = args["cond_scale"]

            # Tonemap
            noise_pred = (cond - uncond)
            noise_pred_vector_magnitude = (torch.linalg.vector_norm(noise_pred, dim=(1)) + 0.0000000001)[:, None]
            noise_pred /= noise_pred_vector_magnitude

            mean = torch.mean(noise_pred_vector_magnitude, dim=(1, 2, 3), keepdim=True)
            std = torch.std(noise_pred_vector_magnitude, dim=(1, 2, 3), keepdim=True)

            top = (std * 3 + mean) * tonemap_multiplier

            # Reinhard
            noise_pred_vector_magnitude *= (1.0 / top)
            new_magnitude = noise_pred_vector_magnitude / (noise_pred_vector_magnitude + 1.0)
            new_magnitude *= top

            # Rescale CFG
            x_cfg = uncond + (noise_pred * new_magnitude * cond_scale)
            ro_pos = torch.std(cond, dim=(1, 2, 3), keepdim=True)
            ro_cfg = torch.std(x_cfg, dim=(1, 2, 3), keepdim=True)

            x_rescaled = x_cfg * (ro_pos / ro_cfg)
            x_final = rescale_multiplier * x_rescaled + (1.0 - rescale_multiplier) * x_cfg

            return x_final

        m = model.clone()
        m.set_model_sampler_cfg_function(tonemap_noise_rescale_cfg)
        return (m, )


NODE_CLASS_MAPPINGS = {
    "TonemapNoiseWithRescaleCFG": TonemapNoiseWithRescaleCFG,
}