Skip to content

Fix dtype error for StableDiffusionXL #9217

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
Aug 20, 2024
Merged

Fix dtype error for StableDiffusionXL #9217

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ea3b2d1
Fix dtype error
leisuzz Aug 19, 2024
3b5bf98
Merge branch 'main' into main
sayakpaul Aug 20, 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
4 changes: 2 additions & 2 deletions examples/text_to_image/train_text_to_image_sdxl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1084,7 +1084,7 @@ def unwrap_model(model):

# Add noise to the model input according to the noise magnitude at each timestep
# (this is the forward diffusion process)
noisy_model_input = noise_scheduler.add_noise(model_input, noise, timesteps)
noisy_model_input = noise_scheduler.add_noise(model_input, noise, timesteps).to(dtype=weight_dtype)
Copy link
Member

@sayakpaul sayakpaul Aug 20, 2024

Choose a reason for hiding this comment

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

Let's do this conditionally if the dtype of noisy_model_input doesn't match with vae.dtype?

Copy link
Contributor Author

@leisuzz leisuzz Aug 20, 2024 •
edited
Loading

Choose a reason for hiding this comment

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

Thanks for your feedback, actually I tested several times, and I found the dtype error actually happens in:

model_pred = unet(
                    noisy_model_input,
                    timesteps,
                    prompt_embeds,
                    added_cond_kwargs=unet_added_conditions,
                    return_dict=False,
                )[0]

So I think it may not affect the vae. Further, the model_pred will be model_pred.float() in the loss part.


# time ids
def compute_time_ids(original_size, crops_coords_top_left):
Expand All @@ -1101,7 +1101,7 @@ def compute_time_ids(original_size, crops_coords_top_left):

# Predict the noise residual
unet_added_conditions = {"time_ids": add_time_ids}
prompt_embeds = batch["prompt_embeds"].to(accelerator.device)
prompt_embeds = batch["prompt_embeds"].to(accelerator.device, dtype=weight_dtype)
pooled_prompt_embeds = batch["pooled_prompt_embeds"].to(accelerator.device)
unet_added_conditions.update({"text_embeds": pooled_prompt_embeds})
model_pred = unet(
Expand Down