Default value docs for trainable_backbone_layers (#4913)

Closes #4905
Co-authored-by: Vasilis Vryniotis <datumbox@users.noreply.github.com>

torchvision/models/detection/faster_rcnn.py

@@ -379,7 +379,8 @@ def fasterrcnn_resnet50_fpn(
         num_classes (int): number of output classes of the model (including the background)
         pretrained_backbone (bool): If True, returns a model with backbone pre-trained on Imagenet
         trainable_backbone_layers (int): number of trainable (not frozen) resnet layers starting from final block.
-            Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable.
+            Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable. If ``None`` is
+            passed (the default) this value is set to 3.
     """
     trainable_backbone_layers = _validate_trainable_layers(
         pretrained or pretrained_backbone, trainable_backbone_layers, 5, 3
@@ -464,7 +465,8 @@ def fasterrcnn_mobilenet_v3_large_320_fpn(
         num_classes (int): number of output classes of the model (including the background)
         pretrained_backbone (bool): If True, returns a model with backbone pre-trained on Imagenet
         trainable_backbone_layers (int): number of trainable (not frozen) resnet layers starting from final block.
-            Valid values are between 0 and 6, with 6 meaning all backbone layers are trainable.
+            Valid values are between 0 and 6, with 6 meaning all backbone layers are trainable. If ``None`` is
+            passed (the default) this value is set to 3.
     """
     weights_name = "fasterrcnn_mobilenet_v3_large_320_fpn_coco"
     defaults = {
@@ -509,7 +511,8 @@ def fasterrcnn_mobilenet_v3_large_fpn(
         num_classes (int): number of output classes of the model (including the background)
         pretrained_backbone (bool): If True, returns a model with backbone pre-trained on Imagenet
         trainable_backbone_layers (int): number of trainable (not frozen) resnet layers starting from final block.
-            Valid values are between 0 and 6, with 6 meaning all backbone layers are trainable.
+            Valid values are between 0 and 6, with 6 meaning all backbone layers are trainable. If ``None`` is
+            passed (the default) this value is set to 3.
     """
     weights_name = "fasterrcnn_mobilenet_v3_large_fpn_coco"
     defaults = {

torchvision/models/detection/keypoint_rcnn.py

@@ -360,7 +360,8 @@ def keypointrcnn_resnet50_fpn(
         num_keypoints (int): number of keypoints, default 17
         pretrained_backbone (bool): If True, returns a model with backbone pre-trained on Imagenet
         trainable_backbone_layers (int): number of trainable (not frozen) resnet layers starting from final block.
-            Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable.
+            Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable. If ``None`` is
+            passed (the default) this value is set to 3.
     """
     trainable_backbone_layers = _validate_trainable_layers(
         pretrained or pretrained_backbone, trainable_backbone_layers, 5, 3

torchvision/models/detection/mask_rcnn.py

@@ -357,7 +357,8 @@ def maskrcnn_resnet50_fpn(
         num_classes (int): number of output classes of the model (including the background)
         pretrained_backbone (bool): If True, returns a model with backbone pre-trained on Imagenet
         trainable_backbone_layers (int): number of trainable (not frozen) resnet layers starting from final block.
-            Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable.
+            Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable. If ``None`` is
+            passed (the default) this value is set to 3.
     """
     trainable_backbone_layers = _validate_trainable_layers(
         pretrained or pretrained_backbone, trainable_backbone_layers, 5, 3

torchvision/models/detection/retinanet.py

@@ -623,7 +623,8 @@ def retinanet_resnet50_fpn(
         num_classes (int): number of output classes of the model (including the background)
         pretrained_backbone (bool): If True, returns a model with backbone pre-trained on Imagenet
         trainable_backbone_layers (int): number of trainable (not frozen) resnet layers starting from final block.
-            Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable.
+            Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable. If ``None`` is
+            passed (the default) this value is set to 3.
     """
     trainable_backbone_layers = _validate_trainable_layers(
         pretrained or pretrained_backbone, trainable_backbone_layers, 5, 3

torchvision/models/detection/ssd.py

@@ -587,7 +587,8 @@ def ssd300_vgg16(
         num_classes (int): number of output classes of the model (including the background)
         pretrained_backbone (bool): If True, returns a model with backbone pre-trained on Imagenet
         trainable_backbone_layers (int): number of trainable (not frozen) resnet layers starting from final block.
-            Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable.
+            Valid values are between 0 and 5, with 5 meaning all backbone layers are trainable. If ``None`` is
+            passed (the default) this value is set to 4.
     """
     if "size" in kwargs:
         warnings.warn("The size of the model is already fixed; ignoring the argument.")

torchvision/models/detection/ssdlite.py

@@ -208,7 +208,8 @@ def ssdlite320_mobilenet_v3_large(
         num_classes (int): number of output classes of the model (including the background)
         pretrained_backbone (bool): If True, returns a model with backbone pre-trained on Imagenet
         trainable_backbone_layers (int): number of trainable (not frozen) resnet layers starting from final block.
-            Valid values are between 0 and 6, with 6 meaning all backbone layers are trainable.
+            Valid values are between 0 and 6, with 6 meaning all backbone layers are trainable. If ``None`` is
+            passed (the default) this value is set to 6.
         norm_layer (callable, optional): Module specifying the normalization layer to use.
     """
     if "size" in kwargs: