wassname/denoising-diffusion-pytorch
1
0
Fork 0
mirror of https://github.com/wassname/denoising-diffusion-pytorch.git synced 2026-09-09 11:21:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

final cleanup

Browse Source
This commit is contained in:
Phil Wang
2022-05-12 13:58:16 -07:00
parent 55c658b967
commit caa5af170d
2 changed files with 2 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
denoising_diffusion_pytorch/learned_gaussian_diffusion.py
+1 -13
View File
@@ -76,18 +76,6 @@ class LearnedGaussianDiffusion(GaussianDiffusion):
assert denoise_fn.out_dim == (denoise_fn.channels * 2), 'dimension out of unet must be twice the number of channels for learned variance - you can also set the `learned_variance` keyword argument on the Unet to be `True`'
self.vb_loss_weight = vb_loss_weight
def q_posterior_mean_variance(self, x_start, x_t, t):
"""
Compute the mean and variance of the diffusion posterior q(x_{t-1} | x_t, x_0)
"""
posterior_mean = (
extract(self.posterior_mean_coef1, t, x_t.shape) * x_start +
extract(self.posterior_mean_coef2, t, x_t.shape) * x_t
)
posterior_variance = extract(self.posterior_variance, t, x_t.shape)
posterior_log_variance_clipped = extract(self.posterior_log_variance_clipped, t, x_t.shape)
return posterior_mean, posterior_variance, posterior_log_variance_clipped
def p_mean_variance(self, *, x, t, clip_denoised, model_output = None):
model_output = default(model_output, lambda: self.denoise_fn(x, t))
pred_noise, var_interp_frac_unnormalized = model_output.chunk(2, dim = 1)
@@ -118,7 +106,7 @@ class LearnedGaussianDiffusion(GaussianDiffusion):
# calculating kl loss for learned variance (interpolation)
true_mean, _, true_log_variance_clipped = self.q_posterior_mean_variance(x_start = x_start, x_t = x_t, t = t)
true_mean, _, true_log_variance_clipped = self.q_posterior(x_start = x_start, x_t = x_t, t = t)
model_mean, _, model_log_variance = self.p_mean_variance(x = x_t, t = t, clip_denoised = clip_denoised, model_output = model_output)
# kl loss with detached model predicted mean, for stability reasons as in paper
setup.py
+1 -1
View File
@@ -3,7 +3,7 @@ from setuptools import setup, find_packages
setup(
name = 'denoising-diffusion-pytorch',
packages = find_packages(),
version = '0.14.2',
version = '0.14.3',
license='MIT',
description = 'Denoising Diffusion Probabilistic Models - Pytorch',
author = 'Phil Wang',