fix: resize transform now preserves aspect ratio (#1804)

parlance-zz · web-flow · commit b693aff7951c · 2022-12-27T15:10:25.000+01:00
diff --git a/examples/text_to_image/train_text_to_image.py b/examples/text_to_image/train_text_to_image.py
@@ -493,7 +493,7 @@ def tokenize_captions(examples, is_train=True):
 
     train_transforms = transforms.Compose(
         [
-            transforms.Resize((args.resolution, args.resolution), interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
             transforms.CenterCrop(args.resolution) if args.center_crop else transforms.RandomCrop(args.resolution),
             transforms.RandomHorizontalFlip() if args.random_flip else transforms.Lambda(lambda x: x),
             transforms.ToTensor(),
diff --git a/examples/text_to_image/train_text_to_image_flax.py b/examples/text_to_image/train_text_to_image_flax.py
@@ -333,7 +333,7 @@ def tokenize_captions(examples, is_train=True):
 
     train_transforms = transforms.Compose(
         [
-            transforms.Resize((args.resolution, args.resolution), interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
             transforms.CenterCrop(args.resolution) if args.center_crop else transforms.RandomCrop(args.resolution),
             transforms.RandomHorizontalFlip() if args.random_flip else transforms.Lambda(lambda x: x),
             transforms.ToTensor(),

Original file line number	Diff line number	Diff line change
`@@ -493,7 +493,7 @@ def tokenize_captions(examples, is_train=True):`
`493`	`493`
`494`	`494`	`train_transforms = transforms.Compose(`
`495`	`495`	`[`
`496`		`- transforms.Resize((args.resolution, args.resolution), interpolation=transforms.InterpolationMode.BILINEAR),`
	`496`	`+ transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),`
`497`	`497`	`transforms.CenterCrop(args.resolution) if args.center_crop else transforms.RandomCrop(args.resolution),`
`498`	`498`	`transforms.RandomHorizontalFlip() if args.random_flip else transforms.Lambda(lambda x: x),`
`499`	`499`	`transforms.ToTensor(),`
Original file line number	Diff line number	Diff line change
`@@ -333,7 +333,7 @@ def tokenize_captions(examples, is_train=True):`
`333`	`333`
`334`	`334`	`train_transforms = transforms.Compose(`
`335`	`335`	`[`
`336`		`- transforms.Resize((args.resolution, args.resolution), interpolation=transforms.InterpolationMode.BILINEAR),`
	`336`	`+ transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),`
`337`	`337`	`transforms.CenterCrop(args.resolution) if args.center_crop else transforms.RandomCrop(args.resolution),`
`338`	`338`	`transforms.RandomHorizontalFlip() if args.random_flip else transforms.Lambda(lambda x: x),`
`339`	`339`	`transforms.ToTensor(),`