Skip to content

feat: allow offset_noise in dreambooth training example #2826

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
sayakpaul merged 1 commit into from
Mar 29, 2023
Merged

feat: allow offset_noise in dreambooth training example #2826

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
17 changes: 16 additions & 1 deletion examples/dreambooth/train_dreambooth.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -417,6 +417,16 @@ def parse_args(input_args=None):
),
)

parser.add_argument(
"--offset_noise",
action="store_true",
default=False,
help=(
"Fine-tuning against a modified noise"
" See: https://www.crosslabs.org//blog/diffusion-with-offset-noise for more information."
),
)

if input_args is not None:
args = parser.parse_args(input_args)
else:
Expand Down Expand Up @@ -943,7 +953,12 @@ def load_model_hook(models, input_dir):
latents = latents * vae.config.scaling_factor

# Sample noise that we'll add to the latents
noise = torch.randn_like(latents)
if args.offset_noise:
noise = torch.randn_like(latents) + 0.1 * torch.randn(
latents.shape[0], latents.shape[1], 1, 1, device=latents.device
)
else:
noise = torch.randn_like(latents)
bsz = latents.shape[0]
# Sample a random timestep for each image
timesteps = torch.randint(0, noise_scheduler.config.num_train_timesteps, (bsz,), device=latents.device)
Expand Down