54 lines
1.8 KiB
Python
54 lines
1.8 KiB
Python
import torch
|
|
from tqdm import trange
|
|
|
|
from comfy.samplers import KSAMPLER
|
|
|
|
|
|
import comfy.k_diffusion.utils as utils
|
|
def to_d(x, sigma, denoised):
|
|
"""Converts a denoiser output to a Karras ODE derivative."""
|
|
return (x - denoised) / utils.append_dims(sigma, x.ndim)
|
|
|
|
|
|
class SamplerInversedEulerNode:
|
|
@classmethod
|
|
def INPUT_TYPES(s):
|
|
return {"required": {
|
|
"refine": ("INT", {"default": 0, "min": 0, "max": 1}),
|
|
}}
|
|
|
|
RETURN_TYPES = ("SAMPLER",)
|
|
CATEGORY = "sampling/custom_sampling/samplers"
|
|
|
|
FUNCTION = "get_sampler"
|
|
|
|
def get_sampler(self, refine):
|
|
@torch.no_grad()
|
|
def sample_inversed_euler(model, x, sigmas, extra_args=None, callback=None, disable=None, s_churn=0., s_tmin=0., s_tmax=float('inf'), s_noise=1.):
|
|
"""Implements Algorithm 2 (Euler steps) from Karras et al. (2022)."""
|
|
extra_args = {} if extra_args is None else extra_args
|
|
s_in = x.new_ones([x.shape[0]])
|
|
for i in trange(1, len(sigmas), disable=disable):
|
|
|
|
dt = sigmas[i] - sigmas[i-1]
|
|
|
|
denoised = model(x, sigmas[i-1] * s_in, **extra_args)
|
|
d = to_d(x, sigmas[i-1], denoised)
|
|
|
|
for _ in range(refine):
|
|
# backward
|
|
x_ = x + d * dt
|
|
# forward
|
|
denoised = model(x_, sigmas[i] * s_in, **extra_args)
|
|
d_ = to_d(x_, sigmas[i], denoised)
|
|
d = d * 0.5 + d_ * 0.5
|
|
|
|
x = x + d * dt
|
|
|
|
if callback is not None:
|
|
callback(
|
|
{'x': x, 'i': i, 'sigma': sigmas[i], 'sigma_hat': sigmas[i], 'denoised': denoised})
|
|
return x
|
|
|
|
ksampler = KSAMPLER(sample_inversed_euler)
|
|
return (ksampler, ) |