Skip to content

Make text-to-image SD LoRA Training Script torch.compile compatible #6555

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 1 commit into from
Jan 15, 2024
Merged

Make text-to-image SD LoRA Training Script torch.compile compatible #6555

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
16 changes: 11 additions & 5 deletions examples/text_to_image/train_text_to_image_lora.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -46,6 +46,7 @@
from diffusers.training_utils import compute_snr
from diffusers.utils import check_min_version, convert_state_dict_to_diffusers, is_wandb_available
from diffusers.utils.import_utils import is_xformers_available
from diffusers.utils.torch_utils import is_compiled_module


# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
Expand Down Expand Up @@ -598,6 +599,11 @@ def tokenize_captions(examples, is_train=True):
]
)

def unwrap_model(model):
model = accelerator.unwrap_model(model)
model = model._orig_mod if is_compiled_module(model) else model
return model

def preprocess_train(examples):
images = [image.convert("RGB") for image in examples[image_column]]
examples["pixel_values"] = [train_transforms(image) for image in images]
Expand Down Expand Up @@ -731,7 +737,7 @@ def collate_fn(examples):
noisy_latents = noise_scheduler.add_noise(latents, noise, timesteps)

# Get the text embedding for conditioning
encoder_hidden_states = text_encoder(batch["input_ids"])[0]
encoder_hidden_states = text_encoder(batch["input_ids"], return_dict=False)[0]

# Get the target for loss depending on the prediction type
if args.prediction_type is not None:
Expand All @@ -746,7 +752,7 @@ def collate_fn(examples):
raise ValueError(f"Unknown prediction type {noise_scheduler.config.prediction_type}")

# Predict the noise residual and compute loss
model_pred = unet(noisy_latents, timesteps, encoder_hidden_states).sample
model_pred = unet(noisy_latents, timesteps, encoder_hidden_states, return_dict=False)[0]
Copy link
Contributor

Choose a reason for hiding this comment

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

why change this?

Copy link
Member

Choose a reason for hiding this comment

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

torch.compile() will throw an error about TensorVariables otherwise.

Copy link
Contributor

Choose a reason for hiding this comment

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

it doesn't here. i just implemented this in ST the other day... didn't change return_dict at all.


if args.snr_gamma is None:
loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
Expand Down Expand Up @@ -811,7 +817,7 @@ def collate_fn(examples):
save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
accelerator.save_state(save_path)

unwrapped_unet = accelerator.unwrap_model(unet)
unwrapped_unet = unwrap_model(unet)
unet_lora_state_dict = convert_state_dict_to_diffusers(
get_peft_model_state_dict(unwrapped_unet)
)
Expand Down Expand Up @@ -839,7 +845,7 @@ def collate_fn(examples):
# create pipeline
pipeline = DiffusionPipeline.from_pretrained(
args.pretrained_model_name_or_path,
unet=accelerator.unwrap_model(unet),
unet=unwrap_model(unet),
revision=args.revision,
variant=args.variant,
torch_dtype=weight_dtype,
Expand Down Expand Up @@ -880,7 +886,7 @@ def collate_fn(examples):
if accelerator.is_main_process:
unet = unet.to(torch.float32)

unwrapped_unet = accelerator.unwrap_model(unet)
unwrapped_unet = unwrap_model(unet)
unet_lora_state_dict = convert_state_dict_to_diffusers(get_peft_model_state_dict(unwrapped_unet))
StableDiffusionPipeline.save_lora_weights(
save_directory=args.output_dir,
Expand Down