Skip to content

Fix train_text_to_image_qat.py #401

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

Closed
AlexanderDokuchaev wants to merge 3 commits into from
Closed

Fix train_text_to_image_qat.py #401

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
ed78c93
train_text_to_image_qat.py
AlexanderDokuchaev Aug 4, 2023
ebcf8ff
black
AlexanderDokuchaev Aug 4, 2023
0dd3dd5
Update examples/openvino/stable-diffusion/requirements.txt
AlexanderDokuchaev Aug 15, 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
2 changes: 1 addition & 1 deletion examples/openvino/stable-diffusion/requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,4 +2,4 @@ accelerate
diffusers
torch~=1.13
Copy link
Contributor

@fxmarty fxmarty Aug 9, 2023

Choose a reason for hiding this comment

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

Is there anything blocking for pytorch 2.0?

Copy link
Contributor Author

@AlexanderDokuchaev AlexanderDokuchaev Aug 15, 2023

Choose a reason for hiding this comment

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

Did not check it.

nncf @ git+https://github.com/openvinotoolkit/nncf.git
tomesd @ git+https://github.com/AlexKoff88/tomesd/tree/openvino
tomesd @ git+https://github.com/AlexKoff88/tomesd.git@openvino
16 changes: 7 additions & 9 deletions examples/openvino/stable-diffusion/train_text_to_image_qat.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -223,7 +223,7 @@ def copy_to(self, parameters: Iterable[torch.nn.Parameter]) -> None:
param.data.copy_(s_param.data)

def to(self, device=None, dtype=None) -> None:
r"""Move internal buffers of the ExponentialMovingAverage to `device`.
"""Move internal buffers of the ExponentialMovingAverage to `device`.

Args:
device: like `device` argument to `torch.Tensor.to`
Expand Down Expand Up @@ -313,7 +313,7 @@ def parse_args():
type=str,
default=None,
choices=["DDIM", "DDPM", "LMSDiscrete"],
help="The noise scheduler for the Diffusion pipiline used for training.",
help="The noise scheduler for the Diffusion pipeline used for training.",
)
parser.add_argument(
"--beta_start",
Expand All @@ -337,7 +337,7 @@ def parse_args():
"--noise_schedule_steps",
type=int,
default=1000,
help=("The noise scheduler max train timestemps"),
help="The noise scheduler max train timestamps",
)
parser.add_argument(
"--center_crop",
Expand Down Expand Up @@ -540,7 +540,7 @@ def parse_args():
type=str,
default="mean_min_max",
choices=["min_max", "mean_min_max", "threesigma"],
help="They way how to estimate activation quantization paramters at the initializatin step before QAT.",
help="They way how to estimate activation quantization parameters at the initialization step before QAT.",
)
parser.add_argument(
"--tune_quantizers_only",
Expand Down Expand Up @@ -775,7 +775,7 @@ def main():
gradient_accumulation_steps=args.gradient_accumulation_steps,
mixed_precision=args.mixed_precision,
log_with=args.report_to,
logging_dir=logging_dir,
project_dir=logging_dir,
)

logging.basicConfig(
Expand Down Expand Up @@ -994,9 +994,7 @@ def collate_fn(examples):
args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
overrode_max_train_steps = True

unet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
unet, optimizer, train_dataloader, lr_scheduler
)
optimizer, train_dataloader, lr_scheduler = accelerator.prepare(optimizer, train_dataloader, lr_scheduler)

weight_dtype = torch.float32
if args.mixed_precision == "fp16":
Expand Down Expand Up @@ -1120,7 +1118,7 @@ def collate_fn(examples):

accelerator.end_training()

# Export optimized pipline to OpenVINO
# Export optimized pipeline to OpenVINO
export_unet = compression_controller.strip(do_copy=False)
export_pipeline = StableDiffusionPipeline(
text_encoder=text_encoder,
Expand Down