mirror of
https://github.com/wassname/denoising-diffusion-pytorch.git
synced 2026-09-12 12:22:11 +08:00
Compare commits
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
eaf9d9fdc4 |
@@ -59,7 +59,6 @@ class ContinuousTimeGaussianDiffusion(nn.Module):
|
|||||||
*,
|
*,
|
||||||
image_size,
|
image_size,
|
||||||
channels = 3,
|
channels = 3,
|
||||||
cond_scale = 500,
|
|
||||||
loss_type = 'l1',
|
loss_type = 'l1',
|
||||||
noise_schedule = 'linear',
|
noise_schedule = 'linear',
|
||||||
num_sample_steps = 500
|
num_sample_steps = 500
|
||||||
@@ -76,7 +75,6 @@ class ContinuousTimeGaussianDiffusion(nn.Module):
|
|||||||
|
|
||||||
# continuous noise schedule related stuff
|
# continuous noise schedule related stuff
|
||||||
|
|
||||||
self.cond_scale = cond_scale # the log(snr) will be scaled by this value
|
|
||||||
self.loss_type = loss_type
|
self.loss_type = loss_type
|
||||||
|
|
||||||
if noise_schedule == 'linear':
|
if noise_schedule == 'linear':
|
||||||
@@ -108,11 +106,6 @@ class ContinuousTimeGaussianDiffusion(nn.Module):
|
|||||||
# todo - derive x_start from the posterior mean and do dynamic thresholding
|
# todo - derive x_start from the posterior mean and do dynamic thresholding
|
||||||
# assumed that is what is going on in Imagen
|
# assumed that is what is going on in Imagen
|
||||||
|
|
||||||
batch = x.shape[0]
|
|
||||||
batch_time = repeat(time, ' -> b', b = batch)
|
|
||||||
|
|
||||||
pred_noise = self.denoise_fn(x, batch_time * self.cond_scale)
|
|
||||||
|
|
||||||
log_snr = self.log_snr(time)
|
log_snr = self.log_snr(time)
|
||||||
log_snr_next = self.log_snr(time_next)
|
log_snr_next = self.log_snr(time_next)
|
||||||
c = -expm1(log_snr - log_snr_next)
|
c = -expm1(log_snr - log_snr_next)
|
||||||
@@ -120,6 +113,9 @@ class ContinuousTimeGaussianDiffusion(nn.Module):
|
|||||||
squared_alpha, squared_alpha_next = log_snr.sigmoid(), log_snr_next.sigmoid()
|
squared_alpha, squared_alpha_next = log_snr.sigmoid(), log_snr_next.sigmoid()
|
||||||
squared_sigma, squared_sigma_next = (-log_snr).sigmoid(), (-log_snr_next).sigmoid()
|
squared_sigma, squared_sigma_next = (-log_snr).sigmoid(), (-log_snr_next).sigmoid()
|
||||||
|
|
||||||
|
batch_log_snr = repeat(log_snr, ' -> b', b = x.shape[0])
|
||||||
|
pred_noise = self.denoise_fn(x, batch_log_snr)
|
||||||
|
|
||||||
model_mean = sqrt(squared_alpha_next / squared_alpha) * (x - c * sqrt(squared_sigma) * pred_noise)
|
model_mean = sqrt(squared_alpha_next / squared_alpha) * (x - c * sqrt(squared_sigma) * pred_noise)
|
||||||
posterior_variance = squared_sigma_next * c
|
posterior_variance = squared_sigma_next * c
|
||||||
|
|
||||||
@@ -151,6 +147,7 @@ class ContinuousTimeGaussianDiffusion(nn.Module):
|
|||||||
times_next = steps[i + 1]
|
times_next = steps[i + 1]
|
||||||
img = self.p_sample(img, times, times_next)
|
img = self.p_sample(img, times, times_next)
|
||||||
|
|
||||||
|
img.clamp_(-1., 1.)
|
||||||
img = unnormalize_to_zero_to_one(img)
|
img = unnormalize_to_zero_to_one(img)
|
||||||
return img
|
return img
|
||||||
|
|
||||||
@@ -180,7 +177,7 @@ class ContinuousTimeGaussianDiffusion(nn.Module):
|
|||||||
|
|
||||||
x, log_snr = self.q_sample(x_start = x_start, times = times, noise = noise)
|
x, log_snr = self.q_sample(x_start = x_start, times = times, noise = noise)
|
||||||
|
|
||||||
model_out = self.denoise_fn(x, log_snr * self.cond_scale)
|
model_out = self.denoise_fn(x, log_snr)
|
||||||
return self.loss_fn(model_out, noise)
|
return self.loss_fn(model_out, noise)
|
||||||
|
|
||||||
def forward(self, img, *args, **kwargs):
|
def forward(self, img, *args, **kwargs):
|
||||||
|
|||||||
@@ -3,7 +3,7 @@ from setuptools import setup, find_packages
|
|||||||
setup(
|
setup(
|
||||||
name = 'denoising-diffusion-pytorch',
|
name = 'denoising-diffusion-pytorch',
|
||||||
packages = find_packages(),
|
packages = find_packages(),
|
||||||
version = '0.16.5',
|
version = '0.16.7',
|
||||||
license='MIT',
|
license='MIT',
|
||||||
description = 'Denoising Diffusion Probabilistic Models - Pytorch',
|
description = 'Denoising Diffusion Probabilistic Models - Pytorch',
|
||||||
author = 'Phil Wang',
|
author = 'Phil Wang',
|
||||||
|
|||||||
Reference in New Issue
Block a user