Skip to content

[Flux] Optimize guidance creation in flux pipeline by moving it outside the loop #9153

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
Aug 16, 2024
Merged

[Flux] Optimize guidance creation in flux pipeline by moving it outside the loop #9153

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
5ce24d0
optimize guidance creation in flux pipeline by moving it outside the …
chengzeyi Aug 12, 2024
e572f9d
use torch.full instead of torch.tensor to create a tensor with a sing…
chengzeyi Aug 12, 2024
4e76315
Merge branch 'main' into optimize_flux
sayakpaul Aug 13, 2024
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
14 changes: 7 additions & 7 deletions src/diffusers/pipelines/flux/pipeline_flux.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -677,6 +677,13 @@ def __call__(
num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)
self._num_timesteps = len(timesteps)

# handle guidance
if self.transformer.config.guidance_embeds:
guidance = torch.full([1], guidance_scale, device=device, dtype=torch.float32)
guidance = guidance.expand(latents.shape[0])
else:
guidance = None
Comment on lines +680 to +685
Copy link
Member

Choose a reason for hiding this comment

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

I like this!


# 6. Denoising loop
with self.progress_bar(total=num_inference_steps) as progress_bar:
for i, t in enumerate(timesteps):
Expand All @@ -686,13 +693,6 @@ def __call__(
# broadcast to batch dimension in a way that's compatible with ONNX/Core ML
timestep = t.expand(latents.shape[0]).to(latents.dtype)

# handle guidance
if self.transformer.config.guidance_embeds:
guidance = torch.tensor([guidance_scale], device=device)
guidance = guidance.expand(latents.shape[0])
else:
guidance = None

noise_pred = self.transformer(
hidden_states=latents,
# YiYi notes: divide it by 1000 for now because we scale it by 1000 in the transforme rmodel (we should not keep it but I want to keep the inputs same for the model for testing)
Expand Down
Loading