Skip to content

Added input pretubation #3292

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 2 commits into from
May 4, 2023
Merged

Added input pretubation #3292

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
bdc0530
Added input pretubation
isamu-isozaki Apr 30, 2023
c9542aa
Fixed spelling
isamu-isozaki Apr 30, 2023
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
11 changes: 9 additions & 2 deletions examples/text_to_image/train_text_to_image.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -112,6 +112,9 @@ def log_validation(vae, text_encoder, tokenizer, unet, args, accelerator, weight

def parse_args():
parser = argparse.ArgumentParser(description="Simple example of a training script.")
parser.add_argument(
"--input_pertubation", type=float, default=0, help="The scale of input pretubation. Recommended 0.1."
)
parser.add_argument(
"--pretrained_model_name_or_path",
type=str,
Expand Down Expand Up @@ -801,15 +804,19 @@ def collate_fn(examples):
noise += args.noise_offset * torch.randn(
(latents.shape[0], latents.shape[1], 1, 1), device=latents.device
)

if args.input_pertubation:
new_noise = noise + args.input_pertubation * torch.randn_like(noise)
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)
timesteps = timesteps.long()

# Add noise to the latents according to the noise magnitude at each timestep
# (this is the forward diffusion process)
noisy_latents = noise_scheduler.add_noise(latents, noise, timesteps)
if args.input_pertubation:
noisy_latents = noise_scheduler.add_noise(latents, new_noise, timesteps)
else:
noisy_latents = noise_scheduler.add_noise(latents, noise, timesteps)

# Get the text embedding for conditioning
encoder_hidden_states = text_encoder(batch["input_ids"])[0]
Expand Down