Skip to content

Add LANCZOS as default interplotation mode. #11463

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 5 commits into from
Apr 30, 2025
Merged

Add LANCZOS as default interplotation mode. #11463

Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
daa4fda
Add LANCZOS as default interplotation mode.
Va16hav07 Apr 30, 2025
4a09f96
Merge branch 'main' into Vaibhav-Improve-Image-Interpolation
Va16hav07 Apr 30, 2025
a35acc9
LANCZOS as default interplotation
Va16hav07 Apr 30, 2025
e94b91d
LANCZOS as default interplotation mode
Va16hav07 Apr 30, 2025
0236a91
Added LANCZOS as default interplotation mode
Va16hav07 Apr 30, 2025
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
32 changes: 28 additions & 4 deletions examples/controlnet/train_controlnet_sdxl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -134,8 +134,19 @@ def log_validation(vae, unet, controlnet, args, accelerator, weight_dtype, step,

for validation_prompt, validation_image in zip(validation_prompts, validation_images):
validation_image = Image.open(validation_image).convert("RGB")
validation_image = validation_image.resize((args.resolution, args.resolution))


# Use the same interpolation mode as in training
if args.interpolation_type.lower() == "lanczos":
interpolation_mode = transforms.InterpolationMode.LANCZOS
else:
interpolation_mode = transforms.InterpolationMode.BILINEAR

transform = transforms.Compose([
transforms.Resize(args.resolution, interpolation=interpolation_mode),
transforms.CenterCrop(args.resolution),
])
validation_image = transform(validation_image)

images = []

for _ in range(args.num_validation_images):
Expand Down Expand Up @@ -587,6 +598,13 @@ def parse_args(input_args=None):
" more information see https://huggingface.co/docs/accelerate/v0.17.0/en/package_reference/accelerator#accelerate.Accelerator"
),
)
parser.add_argument(
"--interpolation_type",
type=str,
default="lanczos",
choices=["lanczos", "bilinear"],
help="The interpolation method to use for resizing images. Choose between 'lanczos' (default) and 'bilinear'.",
)

if input_args is not None:
args = parser.parse_args(input_args)
Expand Down Expand Up @@ -732,9 +750,15 @@ def encode_prompt(prompt_batch, text_encoders, tokenizers, proportion_empty_prom


def prepare_train_dataset(dataset, accelerator):
# Set the interpolation mode based on user preference
if args.interpolation_type.lower() == "lanczos":
interpolation_mode = transforms.InterpolationMode.LANCZOS
else:
interpolation_mode = transforms.InterpolationMode.BILINEAR

image_transforms = transforms.Compose(
[
transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
transforms.Resize(args.resolution, interpolation=interpolation_mode),
transforms.CenterCrop(args.resolution),
transforms.ToTensor(),
transforms.Normalize([0.5], [0.5]),
Expand All @@ -743,7 +767,7 @@ def prepare_train_dataset(dataset, accelerator):

conditioning_image_transforms = transforms.Compose(
[
transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
transforms.Resize(args.resolution, interpolation=interpolation_mode),
transforms.CenterCrop(args.resolution),
transforms.ToTensor(),
]
Expand Down