Add clipUtils (#2473)

* Add clipUtils * Update vit_layers.py * Update vit_layers.py * Update MMDit_block.py * Update clip_utils.py * import fix * Update vit_layers.py * import layer fix * import layer fix * fix imports * import changes * fix failing tests * nit * nit * fix pylint errors * add Copyright * nit --------- Co-authored-by: Divyashree Sreepathihalli <[email protected]>
keras-team · Aug 13, 2024 · e9c0727 · e9c0727
1 parent a452c60
commit e9c0727
Show file tree

Hide file tree

Showing 9 changed files with 365 additions and 58 deletions.
diff --git a/keras_cv/src/layers/vit_layers.py b/keras_cv/src/layers/vit_layers.py
@@ -15,14 +15,14 @@
 import math
 
 import tensorflow as tf
-from keras import layers
-from keras import ops
 
-from keras_cv.api_export import keras_cv_export
+from keras_cv.src.api_export import keras_cv_export
+from keras_cv.src.backend import keras
+from keras_cv.src.backend import ops
 
 
 @keras_cv_export("keras_cv.layers.PatchingAndEmbedding")
-class PatchingAndEmbedding(layers.Layer):
+class PatchingAndEmbedding(keras.layers.Layer):
     """
     Layer to patchify images, prepend a class token, positionally embed and
     create a projection of patches for Vision Transformers
@@ -71,7 +71,7 @@ def __init__(self, project_dim, patch_size, padding="VALID", **kwargs):
                 f"Padding must be either 'SAME' or 'VALID', but {padding} was "
                 "passed."
             )
-        self.projection = layers.Conv2D(
+        self.projection = keras.layers.Conv2D(
             filters=self.project_dim,
             kernel_size=self.patch_size,
             strides=self.patch_size,
@@ -88,7 +88,7 @@ def build(self, input_shape):
             * input_shape[2]
             // self.patch_size
         )
-        self.position_embedding = layers.Embedding(
+        self.position_embedding = keras.layers.Embedding(
             input_dim=self.num_patches + 1, output_dim=self.project_dim
         )
 
@@ -225,32 +225,40 @@ def get_config(self):
 
 
 @keras_cv_export("keras_cv.layers.Unpatching")
-class Unpatching(layers.Layer):
+class Unpatching(keras.layers.Layer):
     """
     Layer to unpatchify image data.
 
-    This layer expects patches sorted by column and reorganizes the patches such that they will each be positioned as a
-    2D shape with some number of channels.
+    This layer expects patches sorted by column and reorganizes the patches
+    such that they will each be positioned as a 2D shape with some
+    number of channels.
 
-    Any necessary padding or truncation will be applied to reach the target shape.
+    Any necessary padding or truncation will be applied to reach the target
+    shape.
 
     Args:
-        target_shape: The target image shape after unpatching, of form [height, width]
+        target_shape: The target image shape after unpatching,
+         of form [height, width]
     """
 
     def __init__(self, target_shape):
         self.target_shape = target_shape
 
     def call(self, patches):
         """
-        Reconstructs an unpatched image from the sequence of column sequence patches.
+        Reconstructs an unpatched image from the sequence of column sequence
+        patches.
 
-        If there are insufficient patches to construct the image of requested dimensions, additional zero-patches will
-        be appended. If excessive patches are provided, unnecessary patches will be truncated from the end.
+        If there are insufficient patches to construct the image of requested
+        dimensions, additional zero-patches will
+        be appended. If excessive patches are provided, unnecessary patches
+        will be truncated from the end.
 
         Args:
-            patches: Patches of images in column sequence (i.e. each patch is vertically oriented relative to the
-                previous patch). Expected shape of [batch_size, patch_num, patch_height, patch_width, channels].
+            patches: Patches of images in column sequence (i.e. each patch
+            is vertically oriented relative to the
+                previous patch). Expected shape of [batch_size, patch_num,
+                 patch_height, patch_width, channels].
 
         Returns:
             Unpatched image: Image reconstructed from the patches,
@@ -284,4 +292,4 @@ def call(self, patches):
         else:
             corrected_patches = patches[:, :required_patches]
 
-        return ops.split(corrected_patches, patches_per_column, axis=1)
+        return ops.split(corrected_patches, patches_per_column, axis=1)
diff --git a/keras_cv/src/models/stable_diffusion_v3/MMDit.py b/keras_cv/src/models/stable_diffusion_v3/MMDit.py
@@ -1,7 +1,21 @@
-from keras_cv.backend import keras
-from keras_cv.layers.vit_layers import PatchingAndEmbedding
-from keras_cv.models.stable_diffusion.v3 import embedding
-from keras_cv.models.stable_diffusion.v3.MMDiT_block import MMDiTBlock
+# Copyright 2024 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 writingf, 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.
+
+from keras_cv.src.backend import keras
+from keras_cv.src.layers.vit_layers import PatchingAndEmbedding
+from keras_cv.src.models.stable_diffusion_v3 import embedding
+from keras_cv.src.models.stable_diffusion_v3.MMDit_block import MMDiTBlock
 
 
 class MMDiT(keras.layers.Layer):

diff --git a/keras_cv/src/models/stable_diffusion_v3/MMDit_block.py b/keras_cv/src/models/stable_diffusion_v3/MMDit_block.py
@@ -1,5 +1,19 @@
-import keras
-from keras_cv.backend import ops
+# Copyright 2024 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 writingf, 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.
+
+from keras_cv.src.backend import keras
+from keras_cv.src.backend import ops
 
 
 class MMDiTSelfAttention(keras.layers.Layer):
@@ -17,7 +31,8 @@ def __init__(
         self.cdense = keras.layers.Dense(key_dim)
 
         if normalization_mode == "rms_normalization":
-            # TODO(varuns1997): Re-Implement RMSNormalization for Keras 2 Compatibility
+            # TODO(varuns1997): Re-Implement RMSNormalization
+            # for Keras 2 Compatibility
             self.query_normalization = keras.layers.LayerNormalization(
                 rms_scaling=True
             )

diff --git a/keras_cv/src/models/stable_diffusion_v3/__init__.py b/keras_cv/src/models/stable_diffusion_v3/__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.