Hotfix/sg 000 fix breaking changes in some imports (#1101)

* Added backward compatibility fixes for imports

* Added backward compatibility fixes for imports

src/super_gradients/training/models/__init__.py

@@ -1,3 +1,5 @@
+import warnings
+
 from .sg_module import SgModule
 
 # Classification models
@@ -33,6 +35,7 @@
     PreActResNet152,
 )
 from super_gradients.training.models.classification_models.resnet import (
+    BasicResNetBlock,
     ResNet,
     ResNet18,
     ResNet34,
@@ -129,6 +132,38 @@
 from super_gradients.common.object_names import Models
 from super_gradients.common.registry.registry import ARCHITECTURES
 
+from super_gradients.training.utils import make_divisible as _make_divisible_current_version
+
+
+def make_deprecated(func, reason):
+    def inner(*args, **kwargs):
+        with warnings.catch_warnings():
+            warnings.simplefilter("once", DeprecationWarning)
+            warnings.warn(reason, category=DeprecationWarning, stacklevel=2)
+        warnings.warn(reason, DeprecationWarning)
+        return func(*args, **kwargs)
+
+    return inner
+
+
+make_divisible = make_deprecated(
+    func=_make_divisible_current_version,
+    reason="You're importing `make_divisible` from `super_gradients.training.models`. This is deprecated since SuperGradients 3.1.0.\n"
+    "Please update your code to import it as follows:\n"
+    "[-] from super_gradients.training.models import make_divisible\n"
+    "[+] from super_gradients.training.utils import make_divisible\n",
+)
+
+
+BasicBlock = make_deprecated(
+    func=BasicResNetBlock,
+    reason="You're importing `BasicBlock` class from `super_gradients.training.models`. This is deprecated since SuperGradients 3.1.0.\n"
+    "This block was renamed to BasicResNetBlock for better clarity.\n"
+    "Please update your code to import it as follows:\n"
+    "[-] from super_gradients.training.models import BasicBlock\n"
+    "[+] from super_gradients.training.models import BasicResNetBlock\n",
+)
+
 __all__ = [
     "SPP",
     "YoloNAS_S",
@@ -293,4 +328,7 @@
     "SegFormerB4",
     "SegFormerB5",
     "DDRNet39Backbone",
+    "make_divisible",
+    "BasicResNetBlock",
+    "BasicBlock",
 ]

src/super_gradients/training/models/classification_models/resnet.py

@@ -22,9 +22,9 @@
 from super_gradients.common.object_names import Models
 
 
-class BasicBlock(nn.Module):
+class BasicResNetBlock(nn.Module):
     def __init__(self, in_planes, planes, stride=1, expansion=1, final_relu=True, droppath_prob=0.0):
-        super(BasicBlock, self).__init__()
+        super(BasicResNetBlock, self).__init__()
         self.expansion = expansion
         self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
         self.bn1 = nn.BatchNorm2d(planes)
@@ -236,7 +236,7 @@ def replace_head(self, new_num_classes=None, new_head=None):
 class ResNet18(ResNet):
     def __init__(self, arch_params, num_classes=None):
         super().__init__(
-            BasicBlock,
+            BasicResNetBlock,
             [2, 2, 2, 2],
             num_classes=num_classes or arch_params.num_classes,
             droppath_prob=get_param(arch_params, "droppath_prob", 0),
@@ -247,14 +247,14 @@ def __init__(self, arch_params, num_classes=None):
 @register_model(Models.RESNET18_CIFAR)
 class ResNet18Cifar(CifarResNet):
     def __init__(self, arch_params, num_classes=None):
-        super().__init__(BasicBlock, [2, 2, 2, 2], num_classes=num_classes or arch_params.num_classes)
+        super().__init__(BasicResNetBlock, [2, 2, 2, 2], num_classes=num_classes or arch_params.num_classes)
 
 
 @register_model(Models.RESNET34)
 class ResNet34(ResNet):
     def __init__(self, arch_params, num_classes=None):
         super().__init__(
-            BasicBlock,
+            BasicResNetBlock,
             [3, 4, 6, 3],
             num_classes=num_classes or arch_params.num_classes,
             droppath_prob=get_param(arch_params, "droppath_prob", 0),
@@ -317,7 +317,7 @@ def __init__(self, arch_params, num_classes=None):
 @register_model(Models.CUSTOM_RESNET_CIFAR)
 class CustomizedResnetCifar(CifarResNet):
     def __init__(self, arch_params, num_classes=None):
-        super().__init__(BasicBlock, arch_params.structure, width_mult=arch_params.width_mult, num_classes=num_classes or arch_params.num_classes)
+        super().__init__(BasicResNetBlock, arch_params.structure, width_mult=arch_params.width_mult, num_classes=num_classes or arch_params.num_classes)
 
 
 @register_model(Models.CUSTOM_RESNET50_CIFAR)
@@ -330,7 +330,7 @@ def __init__(self, arch_params, num_classes=None):
 class CustomizedResnet(ResNet):
     def __init__(self, arch_params, num_classes=None):
         super().__init__(
-            BasicBlock,
+            BasicResNetBlock,
             arch_params.structure,
             width_mult=arch_params.width_mult,
             num_classes=num_classes or arch_params.num_classes,

src/super_gradients/training/models/segmentation_models/ddrnet.py

@@ -5,7 +5,7 @@
 
 from super_gradients.common.registry.registry import register_model
 from super_gradients.common.object_names import Models
-from super_gradients.training.models.classification_models.resnet import BasicBlock, Bottleneck
+from super_gradients.training.models.classification_models.resnet import BasicResNetBlock, Bottleneck
 from super_gradients.training.models.segmentation_models.segmentation_module import SegmentationModule
 from super_gradients.training.utils import get_param, HpmStruct
 
@@ -530,8 +530,8 @@ def __init__(self, arch_params: HpmStruct):
 
 DEFAULT_DDRNET_23_PARAMS = {
     "input_channels": 3,
-    "block": BasicBlock,
-    "skip_block": BasicBlock,
+    "block": BasicResNetBlock,
+    "skip_block": BasicResNetBlock,
     "layer5_block": Bottleneck,
     "layer5_bottleneck_expansion": 2,
     "layers": [2, 2, 2, 2, 1, 2, 2, 1],

src/super_gradients/training/models/segmentation_models/shelfnet.py

@@ -13,7 +13,7 @@
 from super_gradients.training.utils import HpmStruct
 from super_gradients.common.registry.registry import register_model
 from super_gradients.common.object_names import Models
-from super_gradients.training.models.classification_models.resnet import BasicBlock, ResNet, Bottleneck
+from super_gradients.training.models.classification_models.resnet import BasicResNetBlock, ResNet, Bottleneck
 
 
 class FCNHead(nn.Module):
@@ -93,12 +93,12 @@ def forward(self, x):
 
 class ShelfResNetBackBone18(ShelfResNetBackBone):
     def __init__(self, num_classes: int):
-        super().__init__(BasicBlock, [2, 2, 2, 2], num_classes=num_classes)
+        super().__init__(BasicResNetBlock, [2, 2, 2, 2], num_classes=num_classes)
 
 
 class ShelfResNetBackBone34(ShelfResNetBackBone):
     def __init__(self, num_classes: int):
-        super().__init__(BasicBlock, [3, 4, 6, 3], num_classes=num_classes)
+        super().__init__(BasicResNetBlock, [3, 4, 6, 3], num_classes=num_classes)
 
 
 class ShelfResNetBackBone503343(ShelfResNetBackBone):

src/super_gradients/training/utils/__init__.py

@@ -1,4 +1,13 @@
-from super_gradients.training.utils.utils import Timer, HpmStruct, WrappedModel, convert_to_tensor, get_param, tensor_container_to_device, random_seed
+from super_gradients.training.utils.utils import (
+    Timer,
+    HpmStruct,
+    WrappedModel,
+    convert_to_tensor,
+    get_param,
+    tensor_container_to_device,
+    random_seed,
+    make_divisible,
+)
 from super_gradients.training.utils.checkpoint_utils import adapt_state_dict_to_fit_model_layer_names, raise_informative_runtime_error
 from super_gradients.training.utils.version_utils import torch_version_is_greater_or_equal
 from super_gradients.training.utils.config_utils import raise_if_unused_params, warn_if_unused_params
@@ -21,4 +30,5 @@
     "EarlyStop",
     "DEKRPoseEstimationDecodeCallback",
     "DEKRVisualizationCallback",
+    "make_divisible",
 ]

tests/unit_tests/test_deprecations.py

@@ -46,6 +46,24 @@ def test_deprecated_arch_params_top_level_class_via_models_get(self):
         model = models.get("DummyModelV2", arch_params=arch_params, num_classes=80)
         assert isinstance(model, DummyModelV2)
 
+    def test_deprecated_make_divisible(self):
+        try:
+            with self.assertWarns(DeprecationWarning):
+                from super_gradients.training.models import make_divisible  # noqa
+
+                assert make_divisible(1, 1) == 1
+        except ImportError:
+            self.fail("ImportError raised unexpectedly for make_divisible")
+
+    def test_deprecated_BasicBlock(self):
+        try:
+            with self.assertWarns(DeprecationWarning):
+                from super_gradients.training.models import BasicBlock, BasicResNetBlock  # noqa
+
+                assert isinstance(BasicBlock(1, 1, 1), BasicResNetBlock)
+        except ImportError:
+            self.fail("ImportError raised unexpectedly for BasicBlock")
+
 
 if __name__ == "__main__":
     unittest.main()