Throw errors on silent config overwrites from weight meta-data and l…

…egacy builders.

torchvision/prototype/models/_utils.py

@@ -18,3 +18,18 @@ def _deprecated_param(
             raise ValueError("No checkpoint is available for model.")
     else:
         return None
+
+
+def _ovewrite_named_param(kwargs: Dict[str, Any], param: str, new_value: Any) -> None:
+    if param in kwargs:
+        if kwargs[param] != new_value:
+            raise ValueError(f"The parameter {param} expected value {new_value} but got {kwargs[param]} instead.")
+    else:
+        kwargs[param] = new_value
+
+
+def _ovewrite_value_param(param: Any, new_value: Any) -> Any:
+    if param is not None:
+        if param != new_value:
+            raise ValueError(f"The parameter {param} expected value {new_value} but got {param} instead.")
+    return new_value

torchvision/prototype/models/alexnet.py

@@ -7,7 +7,7 @@
 from ...models.alexnet import AlexNet
 from ._api import Weights, WeightEntry
 from ._meta import _IMAGENET_CATEGORIES
-from ._utils import _deprecated_param
+from ._utils import _deprecated_param, _ovewrite_named_param
 
 
 __all__ = ["AlexNet", "AlexNetWeights", "alexnet"]
@@ -34,7 +34,7 @@ def alexnet(weights: Optional[AlexNetWeights] = None, progress: bool = True, **k
     weights = AlexNetWeights.verify(weights)
 
     if weights is not None:
-        kwargs["num_classes"] = len(weights.meta["categories"])
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
 
     model = AlexNet(**kwargs)
 

torchvision/prototype/models/densenet.py

@@ -9,7 +9,7 @@
 from ...models.densenet import DenseNet
 from ._api import Weights, WeightEntry
 from ._meta import _IMAGENET_CATEGORIES
-from ._utils import _deprecated_param
+from ._utils import _deprecated_param, _ovewrite_named_param
 
 
 __all__ = [
@@ -53,7 +53,7 @@ def _densenet(
     **kwargs: Any,
 ) -> DenseNet:
     if weights is not None:
-        kwargs["num_classes"] = len(weights.meta["categories"])
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
 
     model = DenseNet(growth_rate, block_config, num_init_features, **kwargs)
 

torchvision/prototype/models/detection/faster_rcnn.py

@@ -14,7 +14,7 @@
 )
 from .._api import Weights, WeightEntry
 from .._meta import _COCO_CATEGORIES
-from .._utils import _deprecated_param
+from .._utils import _deprecated_param, _ovewrite_value_param
 from ..mobilenetv3 import MobileNetV3LargeWeights, mobilenet_v3_large
 from ..resnet import ResNet50Weights, resnet50
 
@@ -91,7 +91,7 @@ def fasterrcnn_resnet50_fpn(
 
     if weights is not None:
         weights_backbone = None
-        num_classes = len(weights.meta["categories"])
+        num_classes = _ovewrite_value_param(num_classes, len(weights.meta["categories"]))
     elif num_classes is None:
         num_classes = 91
 
@@ -121,7 +121,7 @@ def _fasterrcnn_mobilenet_v3_large_fpn(
 ) -> FasterRCNN:
     if weights is not None:
         weights_backbone = None
-        num_classes = len(weights.meta["categories"])
+        num_classes = _ovewrite_value_param(num_classes, len(weights.meta["categories"]))
     elif num_classes is None:
         num_classes = 91
 

torchvision/prototype/models/detection/keypoint_rcnn.py

@@ -11,7 +11,7 @@
 )
 from .._api import Weights, WeightEntry
 from .._meta import _COCO_PERSON_CATEGORIES, _COCO_PERSON_KEYPOINT_NAMES
-from .._utils import _deprecated_param
+from .._utils import _deprecated_param, _ovewrite_value_param
 from ..resnet import ResNet50Weights, resnet50
 
 
@@ -72,8 +72,8 @@ def keypointrcnn_resnet50_fpn(
 
     if weights is not None:
         weights_backbone = None
-        num_classes = len(weights.meta["categories"])
-        num_keypoints = len(weights.meta["keypoint_names"])
+        num_classes = _ovewrite_value_param(num_classes, len(weights.meta["categories"]))
+        num_keypoints = _ovewrite_value_param(num_keypoints, len(weights.meta["keypoint_names"]))
     else:
         if num_classes is None:
             num_classes = 2

torchvision/prototype/models/detection/mask_rcnn.py

@@ -12,7 +12,7 @@
 )
 from .._api import Weights, WeightEntry
 from .._meta import _COCO_CATEGORIES
-from .._utils import _deprecated_param
+from .._utils import _deprecated_param, _ovewrite_value_param
 from ..resnet import ResNet50Weights, resnet50
 
 
@@ -56,7 +56,7 @@ def maskrcnn_resnet50_fpn(
 
     if weights is not None:
         weights_backbone = None
-        num_classes = len(weights.meta["categories"])
+        num_classes = _ovewrite_value_param(num_classes, len(weights.meta["categories"]))
     elif num_classes is None:
         num_classes = 91
 

torchvision/prototype/models/detection/retinanet.py

@@ -13,7 +13,7 @@
 )
 from .._api import Weights, WeightEntry
 from .._meta import _COCO_CATEGORIES
-from .._utils import _deprecated_param
+from .._utils import _deprecated_param, _ovewrite_value_param
 from ..resnet import ResNet50Weights, resnet50
 
 
@@ -56,7 +56,7 @@ def retinanet_resnet50_fpn(
 
     if weights is not None:
         weights_backbone = None
-        num_classes = len(weights.meta["categories"])
+        num_classes = _ovewrite_value_param(num_classes, len(weights.meta["categories"]))
     elif num_classes is None:
         num_classes = 91
 

torchvision/prototype/models/detection/ssd.py

@@ -12,7 +12,7 @@
 )
 from .._api import Weights, WeightEntry
 from .._meta import _COCO_CATEGORIES
-from .._utils import _deprecated_param
+from .._utils import _deprecated_param, _ovewrite_value_param
 from ..vgg import VGG16Weights, vgg16
 
 
@@ -58,7 +58,7 @@ def ssd300_vgg16(
 
     if weights is not None:
         weights_backbone = None
-        num_classes = len(weights.meta["categories"])
+        num_classes = _ovewrite_value_param(num_classes, len(weights.meta["categories"]))
     elif num_classes is None:
         num_classes = 91
 

torchvision/prototype/models/detection/ssdlite.py

@@ -17,7 +17,7 @@
 )
 from .._api import Weights, WeightEntry
 from .._meta import _COCO_CATEGORIES
-from .._utils import _deprecated_param
+from .._utils import _deprecated_param, _ovewrite_value_param
 from ..mobilenetv3 import MobileNetV3LargeWeights, mobilenet_v3_large
 
 
@@ -64,7 +64,7 @@ def ssdlite320_mobilenet_v3_large(
 
     if weights is not None:
         weights_backbone = None
-        num_classes = len(weights.meta["categories"])
+        num_classes = _ovewrite_value_param(num_classes, len(weights.meta["categories"]))
     elif num_classes is None:
         num_classes = 91
 

torchvision/prototype/models/efficientnet.py

@@ -8,7 +8,7 @@
 from ...models.efficientnet import EfficientNet, MBConvConfig
 from ._api import Weights, WeightEntry
 from ._meta import _IMAGENET_CATEGORIES
-from ._utils import _deprecated_param
+from ._utils import _deprecated_param, _ovewrite_named_param
 
 
 __all__ = [
@@ -41,7 +41,7 @@ def _efficientnet(
     **kwargs: Any,
 ) -> EfficientNet:
     if weights is not None:
-        kwargs["num_classes"] = len(weights.meta["categories"])
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
 
     bneck_conf = partial(MBConvConfig, width_mult=width_mult, depth_mult=depth_mult)
     inverted_residual_setting = [

torchvision/prototype/models/googlenet.py

@@ -8,7 +8,7 @@
 from ...models.googlenet import GoogLeNet, GoogLeNetOutputs, _GoogLeNetOutputs
 from ._api import Weights, WeightEntry
 from ._meta import _IMAGENET_CATEGORIES
-from ._utils import _deprecated_param
+from ._utils import _deprecated_param, _ovewrite_named_param
 
 
 __all__ = ["GoogLeNet", "GoogLeNetOutputs", "_GoogLeNetOutputs", "GoogLeNetWeights", "googlenet"]
@@ -37,10 +37,10 @@ def googlenet(weights: Optional[GoogLeNetWeights] = None, progress: bool = True,
     original_aux_logits = kwargs.get("aux_logits", False)
     if weights is not None:
         if "transform_input" not in kwargs:
-            kwargs["transform_input"] = True
-        kwargs["aux_logits"] = True
-        kwargs["init_weights"] = False
-        kwargs["num_classes"] = len(weights.meta["categories"])
+            _ovewrite_named_param(kwargs, "transform_input", True)
+        _ovewrite_named_param(kwargs, "aux_logits", True)
+        _ovewrite_named_param(kwargs, "init_weights", False)
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
 
     model = GoogLeNet(**kwargs)
 

torchvision/prototype/models/inception.py

@@ -7,7 +7,7 @@
 from ...models.inception import Inception3, InceptionOutputs, _InceptionOutputs
 from ._api import Weights, WeightEntry
 from ._meta import _IMAGENET_CATEGORIES
-from ._utils import _deprecated_param
+from ._utils import _deprecated_param, _ovewrite_named_param
 
 
 __all__ = ["Inception3", "InceptionOutputs", "_InceptionOutputs", "InceptionV3Weights", "inception_v3"]
@@ -36,10 +36,10 @@ def inception_v3(weights: Optional[InceptionV3Weights] = None, progress: bool =
     original_aux_logits = kwargs.get("aux_logits", True)
     if weights is not None:
         if "transform_input" not in kwargs:
-            kwargs["transform_input"] = True
-        kwargs["aux_logits"] = True
-        kwargs["init_weights"] = False
-        kwargs["num_classes"] = len(weights.meta["categories"])
+            _ovewrite_named_param(kwargs, "transform_input", True)
+        _ovewrite_named_param(kwargs, "aux_logits", True)
+        _ovewrite_named_param(kwargs, "init_weights", False)
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
 
     model = Inception3(**kwargs)
 

torchvision/prototype/models/mnasnet.py

@@ -7,7 +7,7 @@
 from ...models.mnasnet import MNASNet
 from ._api import Weights, WeightEntry
 from ._meta import _IMAGENET_CATEGORIES
-from ._utils import _deprecated_param
+from ._utils import _deprecated_param, _ovewrite_named_param
 
 
 __all__ = [
@@ -67,7 +67,7 @@ class MNASNet1_3Weights(Weights):
 
 def _mnasnet(alpha: float, weights: Optional[Weights], progress: bool, **kwargs: Any) -> MNASNet:
     if weights is not None:
-        kwargs["num_classes"] = len(weights.meta["categories"])
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
 
     model = MNASNet(alpha, **kwargs)
 

torchvision/prototype/models/mobilenetv2.py

@@ -7,7 +7,7 @@
 from ...models.mobilenetv2 import MobileNetV2
 from ._api import Weights, WeightEntry
 from ._meta import _IMAGENET_CATEGORIES
-from ._utils import _deprecated_param
+from ._utils import _deprecated_param, _ovewrite_named_param
 
 
 __all__ = ["MobileNetV2", "MobileNetV2Weights", "mobilenet_v2"]
@@ -34,7 +34,7 @@ def mobilenet_v2(weights: Optional[MobileNetV2Weights] = None, progress: bool =
     weights = MobileNetV2Weights.verify(weights)
 
     if weights is not None:
-        kwargs["num_classes"] = len(weights.meta["categories"])
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
 
     model = MobileNetV2(**kwargs)
 

torchvision/prototype/models/mobilenetv3.py

@@ -7,7 +7,7 @@
 from ...models.mobilenetv3 import MobileNetV3, _mobilenet_v3_conf, InvertedResidualConfig
 from ._api import Weights, WeightEntry
 from ._meta import _IMAGENET_CATEGORIES
-from ._utils import _deprecated_param
+from ._utils import _deprecated_param, _ovewrite_named_param
 
 
 __all__ = [
@@ -27,7 +27,7 @@ def _mobilenet_v3(
     **kwargs: Any,
 ) -> MobileNetV3:
     if weights is not None:
-        kwargs["num_classes"] = len(weights.meta["categories"])
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
 
     model = MobileNetV3(inverted_residual_setting, last_channel, **kwargs)
 

torchvision/prototype/models/quantization/googlenet.py

@@ -12,7 +12,7 @@
 )
 from .._api import Weights, WeightEntry
 from .._meta import _IMAGENET_CATEGORIES
-from .._utils import _deprecated_param
+from .._utils import _deprecated_param, _ovewrite_named_param
 from ..googlenet import GoogLeNetWeights
 
 
@@ -60,12 +60,12 @@ def googlenet(
     original_aux_logits = kwargs.get("aux_logits", False)
     if weights is not None:
         if "transform_input" not in kwargs:
-            kwargs["transform_input"] = True
-        kwargs["aux_logits"] = True
-        kwargs["init_weights"] = False
-        kwargs["num_classes"] = len(weights.meta["categories"])
+            _ovewrite_named_param(kwargs, "transform_input", True)
+        _ovewrite_named_param(kwargs, "aux_logits", True)
+        _ovewrite_named_param(kwargs, "init_weights", False)
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
         if "backend" in weights.meta:
-            kwargs["backend"] = weights.meta["backend"]
+            _ovewrite_named_param(kwargs, "backend", weights.meta["backend"])
     backend = kwargs.pop("backend", "fbgemm")
 
     model = QuantizableGoogLeNet(**kwargs)

torchvision/prototype/models/quantization/inception.py

@@ -11,7 +11,7 @@
 )
 from .._api import Weights, WeightEntry
 from .._meta import _IMAGENET_CATEGORIES
-from .._utils import _deprecated_param
+from .._utils import _deprecated_param, _ovewrite_named_param
 from ..inception import InceptionV3Weights
 
 
@@ -59,11 +59,11 @@ def inception_v3(
     original_aux_logits = kwargs.get("aux_logits", False)
     if weights is not None:
         if "transform_input" not in kwargs:
-            kwargs["transform_input"] = True
-        kwargs["aux_logits"] = True
-        kwargs["num_classes"] = len(weights.meta["categories"])
+            _ovewrite_named_param(kwargs, "transform_input", True)
+        _ovewrite_named_param(kwargs, "aux_logits", True)
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
         if "backend" in weights.meta:
-            kwargs["backend"] = weights.meta["backend"]
+            _ovewrite_named_param(kwargs, "backend", weights.meta["backend"])
     backend = kwargs.pop("backend", "fbgemm")
 
     model = QuantizableInception3(**kwargs)

torchvision/prototype/models/quantization/mobilenetv2.py

@@ -12,7 +12,7 @@
 )
 from .._api import Weights, WeightEntry
 from .._meta import _IMAGENET_CATEGORIES
-from .._utils import _deprecated_param
+from .._utils import _deprecated_param, _ovewrite_named_param
 from ..mobilenetv2 import MobileNetV2Weights
 
 
@@ -58,9 +58,9 @@ def mobilenet_v2(
         weights = MobileNetV2Weights.verify(weights)
 
     if weights is not None:
-        kwargs["num_classes"] = len(weights.meta["categories"])
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
         if "backend" in weights.meta:
-            kwargs["backend"] = weights.meta["backend"]
+            _ovewrite_named_param(kwargs, "backend", weights.meta["backend"])
     backend = kwargs.pop("backend", "qnnpack")
 
     model = QuantizableMobileNetV2(block=QuantizableInvertedResidual, **kwargs)

torchvision/prototype/models/quantization/mobilenetv3.py

@@ -13,7 +13,7 @@
 )
 from .._api import Weights, WeightEntry
 from .._meta import _IMAGENET_CATEGORIES
-from .._utils import _deprecated_param
+from .._utils import _deprecated_param, _ovewrite_named_param
 from ..mobilenetv3 import MobileNetV3LargeWeights, _mobilenet_v3_conf
 
 
@@ -33,9 +33,9 @@ def _mobilenet_v3_model(
     **kwargs: Any,
 ) -> QuantizableMobileNetV3:
     if weights is not None:
-        kwargs["num_classes"] = len(weights.meta["categories"])
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
         if "backend" in weights.meta:
-            kwargs["backend"] = weights.meta["backend"]
+            _ovewrite_named_param(kwargs, "backend", weights.meta["backend"])
     backend = kwargs.pop("backend", "qnnpack")
 
     model = QuantizableMobileNetV3(inverted_residual_setting, last_channel, block=QuantizableInvertedResidual, **kwargs)