Revision of dataset normalizations (#898)

pl_bolts/transforms/dataset_normalizations.py

@@ -1,5 +1,6 @@
+from typing import Callable
+
 from pl_bolts.utils import _TORCHVISION_AVAILABLE
-from pl_bolts.utils.stability import under_review
 from pl_bolts.utils.warnings import warn_missing_pkg
 
 if _TORCHVISION_AVAILABLE:
@@ -8,19 +9,17 @@
     warn_missing_pkg("torchvision")
 
 
-@under_review()
-def imagenet_normalization():
+def imagenet_normalization() -> Callable:
     if not _TORCHVISION_AVAILABLE:  # pragma: no cover
         raise ModuleNotFoundError(
             "You want to use `torchvision` which is not installed yet, install it with `pip install torchvision`."
         )
 
-    normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
+    normalize = transforms.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225))
     return normalize
 
 
-@under_review()
-def cifar10_normalization():
+def cifar10_normalization() -> Callable:
     if not _TORCHVISION_AVAILABLE:  # pragma: no cover
         raise ModuleNotFoundError(
             "You want to use `torchvision` which is not installed yet, install it with `pip install torchvision`."
@@ -33,8 +32,7 @@ def cifar10_normalization():
     return normalize
 
 
-@under_review()
-def stl10_normalization():
+def stl10_normalization() -> Callable:
     if not _TORCHVISION_AVAILABLE:  # pragma: no cover
         raise ModuleNotFoundError(
             "You want to use `torchvision` which is not installed yet, install it with `pip install torchvision`."
@@ -44,8 +42,7 @@ def stl10_normalization():
     return normalize
 
 
-@under_review()
-def emnist_normalization(split: str):
+def emnist_normalization(split: str) -> Callable:
     if not _TORCHVISION_AVAILABLE:  # pragma: no cover
         raise ModuleNotFoundError(
             "You want to use `torchvision` which is not installed yet, install it with `pip install torchvision`."

tests/transforms/test_normalizations.py

@@ -0,0 +1,35 @@
+import pytest
+import torch
+from pytorch_lightning import seed_everything
+
+from pl_bolts.transforms.dataset_normalizations import (
+    cifar10_normalization,
+    emnist_normalization,
+    imagenet_normalization,
+    stl10_normalization,
+)
+
+
+@pytest.mark.parametrize(
+    "normalization",
+    [cifar10_normalization, imagenet_normalization, stl10_normalization],
+)
+def test_normalizations(normalization, catch_warnings):
+    """Test normalizations for CIFAR10, ImageNet, STL10."""
+    seed_everything(1234)
+    x = torch.rand(3, 32, 32)
+    assert normalization()(x).shape == (3, 32, 32)
+    assert x.min() >= 0.0
+    assert x.max() <= 1.0
+
+
+@pytest.mark.parametrize(
+    "split",
+    ["balanced", "byclass", "bymerge", "digits", "letters", "mnist"],
+)
+def test_emnist_normalizations(split, catch_warnings):
+    """Test normalizations for each EMNIST dataset split."""
+    x = torch.rand(1, 28, 28)
+    assert emnist_normalization(split)(x).shape == (1, 28, 28)
+    assert x.min() >= 0.0
+    assert x.max() <= 1.0