Skip to content

Revert RePaint scheduler 'fix' #10644

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jan 27, 2025
Merged

Revert RePaint scheduler 'fix' #10644

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 5 additions & 1 deletion src/diffusers/schedulers/scheduling_repaint.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -319,7 +319,11 @@ def step(
prev_unknown_part = alpha_prod_t_prev**0.5 * pred_original_sample + pred_sample_direction + variance

# 8. Algorithm 1 Line 5 https://arxiv.org/pdf/2201.09865.pdf
prev_known_part = (alpha_prod_t_prev**0.5) * original_image + (1 - alpha_prod_t_prev) * noise
# The computation reported in Algorithm 1 Line 5 is incorrect. Line 5 refers to formula (8a) of the same paper,
# which tells to sample from a Gaussian distribution with mean "(alpha_prod_t_prev**0.5) * original_image"
# and variance "(1 - alpha_prod_t_prev)". This means that the standard Gaussian distribution "noise" should be
# scaled by the square root of the variance (as it is done here), however Algorithm 1 Line 5 tells to scale by the variance.
prev_known_part = (alpha_prod_t_prev**0.5) * original_image + ((1 - alpha_prod_t_prev) ** 0.5) * noise

# 9. Algorithm 1 Line 8 https://arxiv.org/pdf/2201.09865.pdf
pred_prev_sample = mask * prev_known_part + (1.0 - mask) * prev_unknown_part
Expand Down