Skip to content

[Training] make checkpointing compatible when using torch.compile (part II) #6511

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
Jan 11, 2024
Merged

[Training] make checkpointing compatible when using torch.compile (part II) #6511

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
79e1489
make checkpointing compatible when using torch.compile.
sayakpaul Jan 10, 2024
e6ea5e1
Merge branch 'main' into torch-compile-compatible-training-ii
sayakpaul Jan 10, 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
24 changes: 15 additions & 9 deletions examples/dreambooth/train_dreambooth_lora_sdxl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,6 +56,7 @@
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 @@ -1007,6 +1008,11 @@ def main(args):
if param.requires_grad:
param.data = param.to(torch.float32)

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

# create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
def save_model_hook(models, weights, output_dir):
if accelerator.is_main_process:
Expand All @@ -1017,13 +1023,13 @@ def save_model_hook(models, weights, output_dir):
text_encoder_two_lora_layers_to_save = None

for model in models:
if isinstance(model, type(accelerator.unwrap_model(unet))):
if isinstance(model, type(unwrap_model(unet))):
unet_lora_layers_to_save = convert_state_dict_to_diffusers(get_peft_model_state_dict(model))
elif isinstance(model, type(accelerator.unwrap_model(text_encoder_one))):
elif isinstance(model, type(unwrap_model(text_encoder_one))):
text_encoder_one_lora_layers_to_save = convert_state_dict_to_diffusers(
get_peft_model_state_dict(model)
)
elif isinstance(model, type(accelerator.unwrap_model(text_encoder_two))):
elif isinstance(model, type(unwrap_model(text_encoder_two))):
text_encoder_two_lora_layers_to_save = convert_state_dict_to_diffusers(
get_peft_model_state_dict(model)
)
Expand All @@ -1048,11 +1054,11 @@ def load_model_hook(models, input_dir):
while len(models) > 0:
model = models.pop()

if isinstance(model, type(accelerator.unwrap_model(unet))):
if isinstance(model, type(unwrap_model(unet))):
unet_ = model
elif isinstance(model, type(accelerator.unwrap_model(text_encoder_one))):
elif isinstance(model, type(unwrap_model(text_encoder_one))):
text_encoder_one_ = model
elif isinstance(model, type(accelerator.unwrap_model(text_encoder_two))):
elif isinstance(model, type(unwrap_model(text_encoder_two))):
text_encoder_two_ = model
else:
raise ValueError(f"unexpected save model: {model.__class__}")
Expand Down Expand Up @@ -1621,16 +1627,16 @@ def compute_text_embeddings(prompt, text_encoders, tokenizers):
# Save the lora layers
accelerator.wait_for_everyone()
if accelerator.is_main_process:
unet = accelerator.unwrap_model(unet)
unet = unwrap_model(unet)
Copy link
Member Author

@sayakpaul sayakpaul Jan 10, 2024

Choose a reason for hiding this comment

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

For intermediate places (such as performing validation inference) where we do use accelerator.unwrap_model() -- it's not an issue as the models are directly used. But here, since we're obtaining the state dicts, we need to get the _orig_mod out in case torch.compile() was called. LMK if it's not clear.

unet = unet.to(torch.float32)
unet_lora_layers = convert_state_dict_to_diffusers(get_peft_model_state_dict(unet))

if args.train_text_encoder:
text_encoder_one = accelerator.unwrap_model(text_encoder_one)
text_encoder_one = unwrap_model(text_encoder_one)
text_encoder_lora_layers = convert_state_dict_to_diffusers(
get_peft_model_state_dict(text_encoder_one.to(torch.float32))
)
text_encoder_two = accelerator.unwrap_model(text_encoder_two)
text_encoder_two = unwrap_model(text_encoder_two)
text_encoder_2_lora_layers = convert_state_dict_to_diffusers(
get_peft_model_state_dict(text_encoder_two.to(torch.float32))
)
Expand Down