[Keras Sprint] Migrate Vision Transformers from Legacy to Backbone API

keras_cv/layers/vit_layers.py

@@ -146,7 +146,7 @@ def call(
             (
                 interpolated_embeddings,
                 class_token,
-            ) = self.__interpolate_positional_embeddings(
+            ) = self.interpolate_positional_embeddings(
                 self.position_embedding(positions),
                 interpolate_width,
                 interpolate_height,
@@ -167,7 +167,7 @@ def call(
             encoded = patches_flattened + self.position_embedding(positions)
         return encoded
 
-    def __interpolate_positional_embeddings(
+    def interpolate_positional_embeddings(
         self, embedding, height, width, patch_size
     ):
         """

keras_cv/layers/vit_layers_test.py

@@ -50,7 +50,7 @@ def test_patch_embedding_interpolation(self):
         (
             output,
             cls,
-        ) = patch_embedding._PatchingAndEmbedding__interpolate_positional_embeddings(  # noqa: E501
+        ) = patch_embedding.interpolate_positional_embeddings(
             positional_embeddings, height=450, width=450, patch_size=12
         )
 
@@ -67,7 +67,7 @@ def test_patch_embedding_interpolation_numerical(self):
         (
             output,
             cls_token,
-        ) = patch_embedding._PatchingAndEmbedding__interpolate_positional_embeddings(  # noqa: E501
+        ) = patch_embedding.interpolate_positional_embeddings(
             positional_embeddings, height=8, width=8, patch_size=2
         )
 

keras_cv/models/__init__.py

@@ -112,6 +112,17 @@
 from keras_cv.models.backbones.resnet_v2.resnet_v2_backbone import (
     ResNetV2Backbone,
 )
+from keras_cv.models.backbones.vit.vit_aliases import ViTB16Backbone
+from keras_cv.models.backbones.vit.vit_aliases import ViTB32Backbone
+from keras_cv.models.backbones.vit.vit_aliases import ViTH16Backbone
+from keras_cv.models.backbones.vit.vit_aliases import ViTH32Backbone
+from keras_cv.models.backbones.vit.vit_aliases import ViTL16Backbone
+from keras_cv.models.backbones.vit.vit_aliases import ViTL32Backbone
+from keras_cv.models.backbones.vit.vit_aliases import ViTS16Backbone
+from keras_cv.models.backbones.vit.vit_aliases import ViTS32Backbone
+from keras_cv.models.backbones.vit.vit_aliases import ViTTiny16Backbone
+from keras_cv.models.backbones.vit.vit_aliases import ViTTiny32Backbone
+from keras_cv.models.backbones.vit.vit_backbone import ViTBackbone
 from keras_cv.models.classification.image_classifier import ImageClassifier
 from keras_cv.models.object_detection.retinanet.retinanet import RetinaNet
 from keras_cv.models.object_detection.yolo_v8.yolo_v8_backbone import (

keras_cv/models/backbones/backbone_presets.py

@@ -22,6 +22,7 @@
 from keras_cv.models.backbones.mobilenet_v3 import mobilenet_v3_backbone_presets
 from keras_cv.models.backbones.resnet_v1 import resnet_v1_backbone_presets
 from keras_cv.models.backbones.resnet_v2 import resnet_v2_backbone_presets
+from keras_cv.models.backbones.vit import vit_backbone_presets
 from keras_cv.models.object_detection.yolo_v8 import yolo_v8_backbone_presets
 
 backbone_presets_no_weights = {
@@ -32,6 +33,7 @@
     **efficientnet_v2_backbone_presets.backbone_presets_no_weights,
     **densenet_backbone_presets.backbone_presets_no_weights,
     **yolo_v8_backbone_presets.backbone_presets_no_weights,
+    **vit_backbone_presets.backbone_presets_no_weights,
 }
 
 backbone_presets_with_weights = {
@@ -42,6 +44,7 @@
     **efficientnet_v2_backbone_presets.backbone_presets_with_weights,
     **densenet_backbone_presets.backbone_presets_with_weights,
     **yolo_v8_backbone_presets.backbone_presets_with_weights,
+    **vit_backbone_presets.backbone_presets_with_weights,
 }
 
 backbone_presets = {

keras_cv/models/backbones/vit/__init__.py

@@ -0,0 +1,13 @@
+# Copyright 2023 The KerasCV Authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     https://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.