Clean up model documentation (#6003)

* Remove old "minimum input size" from docstrings.

* Remove "currently only XYZ weights available"

* Fix description of wide_resnet101_2

* Make display license URLs as links.

* Clarify the order of dims of min_size.

* Remove lengthy keypoint_names from meta-table.

Author: datumbox

docs/source/conf.py

@@ -347,15 +347,19 @@ def inject_weight_metadata(app, what, name, obj, options, lines):
             metrics = meta.pop("metrics", {})
             meta_with_metrics = dict(meta, **metrics)
 
-            meta_with_metrics.pop("categories", None)  # We don't want to document these, they can be too long
+            # We don't want to document these, they can be too long
+            for k in ["categories", "keypoint_names"]:
+                meta_with_metrics.pop(k, None)
 
             custom_docs = meta_with_metrics.pop("_docs", None)  # Custom per-Weights docs
             if custom_docs is not None:
                 lines += [custom_docs, ""]
 
             for k, v in meta_with_metrics.items():
-                if k == "recipe":
+                if k in {"recipe", "license"}:
                     v = f"`link <{v}>`__"
+                elif k == "min_size":
+                    v = f"height={v[0]}, width={v[1]}"
                 table.append((str(k), str(v)))
             table = tabulate(table, tablefmt="rst")
             lines += [".. rst-class:: table-weights"]  # Custom CSS class, see custom_torchvision.css

docs/source/models/mnasnet.rst

@@ -11,8 +11,7 @@ Search for Mobile <https://arxiv.org/pdf/1807.11626.pdf>`__ paper.
 Model builders
 --------------
 
-The following model builders can be used to instanciate an MNASNet model. Currently
-only ``mnasnet0_5`` and ``mnasnet1_0`` can be instantiated with pre-trained weights. 
+The following model builders can be used to instanciate an MNASNet model.
 All the model builders internally rely on the
 ``torchvision.models.mnasnet.MNASNet`` base class. Please refer to the `source
 code

torchvision/models/alexnet.py

@@ -74,8 +74,6 @@ class AlexNet_Weights(WeightsEnum):
 def alexnet(*, weights: Optional[AlexNet_Weights] = None, progress: bool = True, **kwargs: Any) -> AlexNet:
     """AlexNet model architecture from `One weird trick for parallelizing convolutional neural networks <https://arxiv.org/abs/1404.5997>`__.
 
-    The required minimum input size of the model is 63x63.
-
     .. note::
         AlexNet was originally introduced in the `ImageNet Classification with
         Deep Convolutional Neural Networks

torchvision/models/densenet.py

@@ -332,7 +332,6 @@ class DenseNet201_Weights(WeightsEnum):
 def densenet121(*, weights: Optional[DenseNet121_Weights] = None, progress: bool = True, **kwargs: Any) -> DenseNet:
     r"""Densenet-121 model from
     `Densely Connected Convolutional Networks <https://arxiv.org/abs/1608.06993>`_.
-    The required minimum input size of the model is 29x29.
 
     Args:
         weights (:class:`~torchvision.models.DenseNet121_Weights`, optional): The
@@ -358,7 +357,6 @@ def densenet121(*, weights: Optional[DenseNet121_Weights] = None, progress: bool
 def densenet161(*, weights: Optional[DenseNet161_Weights] = None, progress: bool = True, **kwargs: Any) -> DenseNet:
     r"""Densenet-161 model from
     `Densely Connected Convolutional Networks <https://arxiv.org/abs/1608.06993>`_.
-    The required minimum input size of the model is 29x29.
 
     Args:
         weights (:class:`~torchvision.models.DenseNet161_Weights`, optional): The
@@ -384,7 +382,6 @@ def densenet161(*, weights: Optional[DenseNet161_Weights] = None, progress: bool
 def densenet169(*, weights: Optional[DenseNet169_Weights] = None, progress: bool = True, **kwargs: Any) -> DenseNet:
     r"""Densenet-169 model from
     `Densely Connected Convolutional Networks <https://arxiv.org/abs/1608.06993>`_.
-    The required minimum input size of the model is 29x29.
 
     Args:
         weights (:class:`~torchvision.models.DenseNet169_Weights`, optional): The
@@ -410,7 +407,6 @@ def densenet169(*, weights: Optional[DenseNet169_Weights] = None, progress: bool
 def densenet201(*, weights: Optional[DenseNet201_Weights] = None, progress: bool = True, **kwargs: Any) -> DenseNet:
     r"""Densenet-201 model from
     `Densely Connected Convolutional Networks <https://arxiv.org/abs/1608.06993>`_.
-    The required minimum input size of the model is 29x29.
 
     Args:
         weights (:class:`~torchvision.models.DenseNet201_Weights`, optional): The

torchvision/models/googlenet.py

@@ -298,8 +298,6 @@ def googlenet(*, weights: Optional[GoogLeNet_Weights] = None, progress: bool = T
     """GoogLeNet (Inception v1) model architecture from
     `Going Deeper with Convolutions <http://arxiv.org/abs/1409.4842>`_.
 
-    The required minimum input size of the model is 15x15.
-
     Args:
         weights (:class:`~torchvision.models.GoogLeNet_Weights`, optional): The
             pretrained weights for the model. See

torchvision/models/inception.py

@@ -430,7 +430,6 @@ def inception_v3(*, weights: Optional[Inception_V3_Weights] = None, progress: bo
     """
     Inception v3 model architecture from
     `Rethinking the Inception Architecture for Computer Vision <http://arxiv.org/abs/1512.00567>`_.
-    The required minimum input size of the model is 75x75.
 
     .. note::
         **Important**: In contrast to the other models the inception_v3 expects tensors with a size of

torchvision/models/quantization/googlenet.py

@@ -147,8 +147,6 @@ def googlenet(
     weights. Quantized models only support inference and run on CPUs.
     GPU inference is not yet supported
 
-    The required minimum input size of the model is 15x15.
-
     Args:
         weights (:class:`~torchvision.models.quantization.GoogLeNet_QuantizedWeights` or :class:`~torchvision.models.GoogLeNet_Weights`, optional): The
             pretrained weights for the model. See

torchvision/models/resnet.py

@@ -831,8 +831,8 @@ def wide_resnet101_2(
 
     The model is the same as ResNet except for the bottleneck number of channels
     which is twice larger in every block. The number of channels in outer 1x1
-    convolutions is the same, e.g. last block in ResNet-50 has 2048-512-2048
-    channels, and in Wide ResNet-50-2 has 2048-1024-2048.
+    convolutions is the same, e.g. last block in ResNet-101 has 2048-512-2048
+    channels, and in Wide ResNet-101-2 has 2048-1024-2048.
 
     Args:
         weights (:class:`~torchvision.models.Wide_ResNet101_2_Weights`, optional): The

torchvision/models/squeezenet.py

@@ -162,8 +162,6 @@ def squeezenet1_0(
     accuracy with 50x fewer parameters and <0.5MB model size
     <https://arxiv.org/abs/1602.07360>`_ paper.
 
-    The required minimum input size of the model is 21x21.
-
     Args:
         weights (:class:`~torchvision.models.SqueezeNet1_0_Weights`, optional): The
             pretrained weights to use. See
@@ -193,7 +191,6 @@ def squeezenet1_1(
 
     SqueezeNet 1.1 has 2.4x less computation and slightly fewer parameters
     than SqueezeNet 1.0, without sacrificing accuracy.
-    The required minimum input size of the model is 17x17.
 
     Args:
         weights (:class:`~torchvision.models.SqueezeNet1_1_Weights`, optional): The