Mirrors
/
diffusers
mirror of https://github.com/ShivamShrirao/diffusers.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Update README.md

This commit is contained in:
Patrick von Platen 2022-06-09 18:31:37 +02:00 committed by GitHub
parent 2852c80540
commit 1122c7079a
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
README.md
View File

@ -50,8 +50,8 @@ for t in reversed(range(len(scheduler))):
pred_noise_t = self.unet(image, t) pred_noise_t = self.unet(image, t)
# 2. compute alphas, betas # 2. compute alphas, betas
alpha_prod_t = self.noise_scheduler.get_alpha_prod(t) alpha_prod_t = scheduler.get_alpha_prod(t)
alpha_prod_t_prev = self.noise_scheduler.get_alpha_prod(t - 1) alpha_prod_t_prev = scheduler.get_alpha_prod(t - 1)
beta_prod_t = 1 - alpha_prod_t beta_prod_t = 1 - alpha_prod_t
beta_prod_t_prev = 1 - alpha_prod_t_prev beta_prod_t_prev = 1 - alpha_prod_t_prev
@ -65,8 +65,8 @@ for t in reversed(range(len(scheduler))):
# Third: Compute coefficients for pred_original_image x_0 and current image x_t # Third: Compute coefficients for pred_original_image x_0 and current image x_t
# See formula (7) from https://arxiv.org/pdf/2006.11239.pdf # See formula (7) from https://arxiv.org/pdf/2006.11239.pdf
pred_original_image_coeff = (alpha_prod_t_prev.sqrt() * self.noise_scheduler.get_beta(t)) / beta_prod_t pred_original_image_coeff = (alpha_prod_t_prev.sqrt() * scheduler.get_beta(t)) / beta_prod_t
current_image_coeff = self.noise_scheduler.get_alpha(t).sqrt() * beta_prod_t_prev / beta_prod_t current_image_coeff = scheduler.get_alpha(t).sqrt() * beta_prod_t_prev / beta_prod_t
# Fourth: Compute predicted previous image µ_t # Fourth: Compute predicted previous image µ_t
# See formula (7) from https://arxiv.org/pdf/2006.11239.pdf # See formula (7) from https://arxiv.org/pdf/2006.11239.pdf
pred_prev_image = pred_original_image_coeff * pred_original_image + current_image_coeff * image pred_prev_image = pred_original_image_coeff * pred_original_image + current_image_coeff * image
@ -76,7 +76,7 @@ for t in reversed(range(len(scheduler))):
# x_{t-1} ~ N(pred_prev_image, variance) == add variane to pred_image # x_{t-1} ~ N(pred_prev_image, variance) == add variane to pred_image
if t > 0: if t > 0:
variance = (1 - alpha_prod_t_prev) / (1 - alpha_prod_t) * self.noise_scheduler.get_beta(t).sqrt() variance = (1 - alpha_prod_t_prev) / (1 - alpha_prod_t) * self.noise_scheduler.get_beta(t).sqrt()
noise = self.noise_scheduler.sample_noise(image.shape, device=image.device, generator=generator) noise = scheduler.sample_noise(image.shape, device=image.device, generator=generator)
prev_image = pred_prev_image + variance * noise prev_image = pred_prev_image + variance * noise
else: else:
prev_image = pred_prev_image prev_image = pred_prev_image