laksjdjf

import comfy
from comfy.samplers import KSAMPLER
import torch
from comfy.k_diffusion.sampling import default_noise_sampler, get_ancestral_step, to_d, BrownianTreeNoiseSampler
from tqdm.auto import trange

@torch.no_grad()
def sample_euler_ancestral(model, x, sigmas, extra_args=None, callback=None, disable=None, eta=1., s_noise=1., noise_sampler=None, upscale_ratio=2.0, start_step=5, end_step=15, upscale_n_step=3):
    """Ancestral sampling with Euler method steps."""
    extra_args = {} if extra_args is None else extra_args

laksjdjf

================================================================================================================================================================
Layer (type (var_name))                                      Input Shape               Output Shape              Param #                   Kernel Shape
================================================================================================================================================================
SD3Transformer2DModel (SD3Transformer2DModel)                --                        [1, 16, 128, 128]         --                        --
├─PatchEmbed (pos_embed)                                     [1, 16, 128, 128]         [1, 4096, 1536]           --                        --
│    └─Conv2d (proj)                                         [1, 16, 128, 128]         [1, 1536, 64, 64]         99,840                    [2, 2]
├─CombinedTimestepTextProjEmbeddings (time_text_embed)       [1]                       [1, 1536]                 --     

laksjdjf

import numpy as np
import matplotlib.pyplot as plt

def sigmoid(x):
    return 1 / (1 + np.exp(-x))

def inverse_sigmoid(y):
    return np.log(y / (1 - y))

# 逆シグモイド関数の微分