compatibility fixes for new albumentations version

Author: AdeelH

rastervision_pytorch_learner/rastervision/pytorch_learner/dataset/dataset.py

@@ -163,7 +163,7 @@ def append_resize_transform(
             self, transform: A.BasicTransform | None,
             out_size: tuple[PosInt, PosInt]) -> A.Resize | A.Compose:
         """Get transform to use for resizing windows to out_size."""
-        resize_tf = A.Resize(*out_size, always_apply=True)
+        resize_tf = A.Resize(*out_size, p=1.)
         if transform is None:
             transform = resize_tf
         else:

rastervision_pytorch_learner/rastervision/pytorch_learner/dataset/object_detection_dataset.py

@@ -235,7 +235,7 @@ def __init__(self, *args, **kwargs):
 
     def append_resize_transform(self, transform: A.BasicTransform,
                                 out_size: tuple[int, int]) -> A.BasicTransform:
-        resize_tf = A.Resize(*out_size, always_apply=True)
+        resize_tf = A.Resize(*out_size, p=1.)
         if transform is None:
             transform = resize_tf
         else:

rastervision_pytorch_learner/rastervision/pytorch_learner/utils/utils.py

@@ -222,7 +222,6 @@ def __init__(
             min_val=0.0,
             max_val=1.0,
             dtype=cv2.CV_32F,
-            always_apply=False,
             p=1.0,
     ):
         """Constructor.
@@ -232,7 +231,7 @@ def __init__(
             max_val: the maximum value that output should have
             dtype: the dtype of output image
         """
-        super(MinMaxNormalize, self).__init__(always_apply, p)
+        super(MinMaxNormalize, self).__init__(p)
         self.min_val = min_val
         self.max_val = max_val
         self.dtype = dtype

tests/data_files/lambda_transforms.py

@@ -18,6 +18,6 @@ def swap(image: 'np.ndarray', **kwargs) -> 'np.ndarray':
 
 
 lambda_transforms = {
-    'ndvi': A.Lambda(name='ndvi', image=ndvi, always_apply=True),
-    'swap': A.Lambda(name='swap', image=swap, always_apply=True)
+    'ndvi': A.Lambda(name='ndvi', image=ndvi, p=1.),
+    'swap': A.Lambda(name='swap', image=swap, p=1.)
 }