Skip to content

[fix bug] PixArt inference_steps=1 #11079

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 3 commits into from
Mar 20, 2025
Merged

[fix bug] PixArt inference_steps=1 #11079

Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2dae26c
fix bug when pixart-dmd inference with `num_inference_steps=1`
lawrence-cj Mar 17, 2025
ae78a65
use return_dict=False and return [1] element for 1-step pixart model,…
lawrence-cj Mar 20, 2025
c577fcf
Merge branch 'main' into main
lawrence-cj Mar 20, 2025
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
8 changes: 6 additions & 2 deletions src/diffusers/pipelines/pixart_alpha/pipeline_pixart_alpha.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -941,8 +941,12 @@ def __call__(

# compute previous image: x_t -> x_t-1
if num_inference_steps == 1:
# For DMD one step sampling: https://arxiv.org/abs/2311.18828
latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).pred_original_sample
try:
# For LCM one step sampling
latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).denoised
except:
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
except:
except Exception:

At minimum due to ruff. What's the exception type raised here? Is there anything else we can check instead of relying on try/except?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This one is due to the Output difference between

diffusers/src/diffusers/schedulers/scheduling_lcm.py

Lines 48 to 49 in 1001425

prev_sample: torch.Tensor
denoised: Optional[torch.Tensor] = None

and

diffusers/src/diffusers/schedulers/scheduling_ddim.py

Lines 46 to 47 in 1001425

prev_sample: torch.Tensor
pred_original_sample: Optional[torch.Tensor] = None

What do you think is better to write here?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

gentle ping @hlky

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

With return_dict=False a tuple is returned, both denoised and pred_original_sample are at index 1 so we can do

latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[1]

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That makes sense. updated.

# For DMD one step sampling: https://arxiv.org/abs/2311.18828
latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).pred_original_sample
else:
latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]

Expand Down
Loading