Skip to content

feat(training-utils): support device and dtype params in compute_density_for_timestep_sampling #10699

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 4 commits into from
Feb 1, 2025
Merged

feat(training-utils): support device and dtype params in compute_density_for_timestep_sampling #10699

Changes from 2 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
3667b3b
feat(training-utils): support device and dtype params in compute_dens…
badayvedat Feb 1, 2025
8c8ada4
chore: update type hint
badayvedat Feb 1, 2025
ba03950
refactor: use union for type hint
badayvedat Feb 1, 2025
b92d43f
Merge branch 'main' into feat/add-device-tpype
sayakpaul Feb 1, 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
15 changes: 10 additions & 5 deletions src/diffusers/training_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -248,7 +248,13 @@ def _set_state_dict_into_text_encoder(


def compute_density_for_timestep_sampling(
weighting_scheme: str, batch_size: int, logit_mean: float = None, logit_std: float = None, mode_scale: float = None
weighting_scheme: str,
batch_size: int,
logit_mean: float = None,
logit_std: float = None,
mode_scale: float = None,
device: torch.device | str = "cpu",
Copy link
Contributor

Choose a reason for hiding this comment

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

Could we use Union here instead?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

sure! and i know the same change because i was trying to minimize the duplications in finetrainers 😄

generator: Optional[torch.Generator] = None,
):
"""
Compute the density for sampling the timesteps when doing SD3 training.
Expand All @@ -258,14 +264,13 @@ def compute_density_for_timestep_sampling(
SD3 paper reference: https://arxiv.org/abs/2403.03206v1.
"""
if weighting_scheme == "logit_normal":
# See 3.1 in the SD3 paper ($rf/lognorm(0.00,1.00)$).
u = torch.normal(mean=logit_mean, std=logit_std, size=(batch_size,), device="cpu")
u = torch.normal(mean=logit_mean, std=logit_std, size=(batch_size,), device=device, generator=generator)
u = torch.nn.functional.sigmoid(u)
elif weighting_scheme == "mode":
u = torch.rand(size=(batch_size,), device="cpu")
u = torch.rand(size=(batch_size,), device=device, generator=generator)
u = 1 - u - mode_scale * (torch.cos(math.pi * u / 2) ** 2 - 1 + u)
else:
u = torch.rand(size=(batch_size,), device="cpu")
u = torch.rand(size=(batch_size,), device=device, generator=generator)
return u


Expand Down
Loading