tf-keras-nightly 2.17.0.dev2024031909__tar.gz → 2.19.0.dev2025011410__tar.gz

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.1
+Metadata-Version: 2.2
 Name: tf_keras-nightly
-Version: 2.17.0.dev2024031909
+Version: 2.19.0.dev2025011410
 Summary: Deep learning for humans.
 Home-page: https://keras.io/
 Download-URL: https://github.com/keras-team/tf-keras/tags
@@ -26,7 +26,18 @@ Classifier: Topic :: Software Development
 Classifier: Topic :: Software Development :: Libraries
 Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Requires-Python: >=3.9
-Requires-Dist: tf-nightly~=2.17.0.dev
+Requires-Dist: tf-nightly~=2.19.0.dev
+Dynamic: author
+Dynamic: author-email
+Dynamic: classifier
+Dynamic: description
+Dynamic: download-url
+Dynamic: home-page
+Dynamic: keywords
+Dynamic: license
+Dynamic: requires-dist
+Dynamic: requires-python
+Dynamic: summary
 
 TF-Keras is a deep learning API written in Python,
 running on top of the machine learning platform TensorFlow.

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/setup.py

@@ -38,7 +38,7 @@ with open(os.path.abspath(__file__)) as f:
         )
 
 # pin version to that of tensorflow or tf_nightly.
-version = "2.17.0.dev2024031909".lower()
+version = "2.19.0.dev2025011410".lower()
 major_version, minor_version, *_ = version.split(".")
 next_minor_version = int(minor_version) + 1
 if "nightly" in "tf_keras-nightly":
@@ -61,7 +61,7 @@ setuptools.setup(
     name="tf_keras-nightly",
     # Version strings with `-` characters are semver compatible,
     # but incompatible with pip. For pip, we will remove all `-`` characters.
-    version="2.17.0.dev2024031909",
+    version="2.19.0.dev2025011410",
     description="Deep learning for humans.",
     long_description=DESCRIPTION,
     url="https://keras.io/",

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/__init__.py

@@ -27,4 +27,4 @@ from tf_keras.src.engine.sequential import Sequential
 from tf_keras.src.engine.training import Model
 
 
-__version__ = "2.17.0.dev2024031909"
+__version__ = "2.19.0.dev2025011410"

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/__init__.py

@@ -35,7 +35,7 @@ from tf_keras.src.testing_infra import test_utils
 from tensorflow.python import tf2
 from tensorflow.python.util.tf_export import keras_export
 
-__version__ = "2.17.0"
+__version__ = "2.19.0"
 
 keras_export("keras.__version__").export_constant(__name__, "__version__")
 

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/backend.py

@@ -2029,7 +2029,7 @@ class RandomGenerator(tf.__internal__.tracking.AutoTrackable):
         if user_specified_seed is not None:
             return user_specified_seed
         elif getattr(_SEED_GENERATOR, "generator", None):
-            return _SEED_GENERATOR.generator.randint(1, 1e9)
+            return _SEED_GENERATOR.generator.randint(1, int(1e9))
         else:
             return random.randint(1, int(1e9))
 

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/callbacks.py

@@ -1423,20 +1423,20 @@ class ModelCheckpoint(Callback):
         if mode == "min":
             self.monitor_op = np.less
             if self.best is None:
-                self.best = np.Inf
+                self.best = np.inf
         elif mode == "max":
             self.monitor_op = np.greater
             if self.best is None:
-                self.best = -np.Inf
+                self.best = -np.inf
         else:
             if "acc" in self.monitor or self.monitor.startswith("fmeasure"):
                 self.monitor_op = np.greater
                 if self.best is None:
-                    self.best = -np.Inf
+                    self.best = -np.inf
             else:
                 self.monitor_op = np.less
                 if self.best is None:
-                    self.best = np.Inf
+                    self.best = np.inf
 
         if self.save_freq != "epoch" and not isinstance(self.save_freq, int):
             raise ValueError(
@@ -1903,6 +1903,23 @@ class BackupAndRestore(Callback):
                 "only supports empty strategy, "
                 "MirroredStrategy, MultiWorkerMirroredStrategy and TPUStrategy."
             )
+
+        # Re-initialize the optimizer.
+        if self.model.built:
+            if (
+                self.model.optimizer is not None
+                and callable(getattr(self.model.optimizer, "build", None))
+                and not getattr(self.model.optimizer, "_built", False)
+            ):
+                self.model.optimizer.build(self.model.trainable_variables)
+        else:
+            logging.warning(
+                "To use the BackupAndRestore callback, "
+                "you model must be built before you call `fit()`. "
+                f"Model {self.model} is unbuilt. You can build it "
+                "beforehand by calling it on a batch of data."
+            )
+
         self.model._training_state = worker_training_state.WorkerTrainingState(
             self.model,
             self.backup_dir,
@@ -2095,7 +2112,7 @@ class EarlyStopping(Callback):
         # Allow instances to be re-used
         self.wait = 0
         self.stopped_epoch = 0
-        self.best = np.Inf if self.monitor_op == np.less else -np.Inf
+        self.best = np.inf if self.monitor_op == np.less else -np.inf
         self.best_weights = None
         self.best_epoch = 0
 
@@ -3098,10 +3115,10 @@ class ReduceLROnPlateau(Callback):
             self.mode == "auto" and "acc" not in self.monitor
         ):
             self.monitor_op = lambda a, b: np.less(a, b - self.min_delta)
-            self.best = np.Inf
+            self.best = np.inf
         else:
             self.monitor_op = lambda a, b: np.greater(a, b + self.min_delta)
-            self.best = -np.Inf
+            self.best = -np.inf
         self.cooldown_counter = 0
         self.wait = 0
 

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/datasets/boston_housing.py

@@ -14,6 +14,8 @@
 # ==============================================================================
 """Boston housing price regression dataset."""
 
+import os
+
 import numpy as np
 
 from tf_keras.src.utils.data_utils import get_file
@@ -23,7 +25,9 @@ from tensorflow.python.util.tf_export import keras_export
 
 
 @keras_export("keras.datasets.boston_housing.load_data")
-def load_data(path="boston_housing.npz", test_split=0.2, seed=113):
+def load_data(
+    path="boston_housing.npz", test_split=0.2, seed=113, cache_dir=None
+):
     """Loads the Boston Housing dataset.
 
     This is a dataset taken from the StatLib library which is maintained at
@@ -43,11 +47,12 @@ def load_data(path="boston_housing.npz", test_split=0.2, seed=113):
     [StatLib website](http://lib.stat.cmu.edu/datasets/boston).
 
     Args:
-      path: path where to cache the dataset locally
-          (relative to `~/.keras/datasets`).
+      path: path where to cache the dataset locally (relative to
+        `~/.keras/datasets`).
       test_split: fraction of the data to reserve as test set.
-      seed: Random seed for shuffling the data
-          before computing the test split.
+      seed: Random seed for shuffling the data before computing the test split.
+      cache_dir: directory where to cache the dataset locally. When None,
+        defaults to `~/.keras/datasets`.
 
     Returns:
       Tuple of Numpy arrays: `(x_train, y_train), (x_test, y_test)`.
@@ -64,12 +69,16 @@ def load_data(path="boston_housing.npz", test_split=0.2, seed=113):
     origin_folder = (
         "https://storage.googleapis.com/tensorflow/tf-keras-datasets/"
     )
+    if cache_dir:
+        cache_dir = os.path.expanduser(cache_dir)
+        os.makedirs(cache_dir, exist_ok=True)
     path = get_file(
         path,
         origin=origin_folder + "boston_housing.npz",
         file_hash=(  # noqa: E501
             "f553886a1f8d56431e820c5b82552d9d95cfcb96d1e678153f8839538947dff5"
         ),
+        cache_dir=cache_dir,
     )
     with np.load(path, allow_pickle=True) as f:
         x = f["x"]

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/datasets/cifar10.py

@@ -27,7 +27,7 @@ from tensorflow.python.util.tf_export import keras_export
 
 
 @keras_export("keras.datasets.cifar10.load_data")
-def load_data():
+def load_data(cache_dir=None):
     """Loads the CIFAR10 dataset.
 
     This is a dataset of 50,000 32x32 color training images and 10,000 test
@@ -49,6 +49,10 @@ def load_data():
     |   8   | ship        |
     |   9   | truck       |
 
+    Args:
+      cache_dir: directory where to cache the dataset locally. When None,
+        defaults to `~/.keras/datasets`.
+
     Returns:
       Tuple of NumPy arrays: `(x_train, y_train), (x_test, y_test)`.
 
@@ -78,6 +82,9 @@ def load_data():
     """
     dirname = "cifar-10-batches-py"
     origin = "https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz"
+    if cache_dir:
+        cache_dir = os.path.expanduser(cache_dir)
+        os.makedirs(cache_dir, exist_ok=True)
     path = get_file(
         dirname,
         origin=origin,
@@ -85,6 +92,7 @@ def load_data():
         file_hash=(  # noqa: E501
             "6d958be074577803d12ecdefd02955f39262c83c16fe9348329d7fe0b5c001ce"
         ),
+        cache_dir=cache_dir,
     )
 
     num_train_samples = 50000

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/datasets/cifar100.py

@@ -27,7 +27,7 @@ from tensorflow.python.util.tf_export import keras_export
 
 
 @keras_export("keras.datasets.cifar100.load_data")
-def load_data(label_mode="fine"):
+def load_data(label_mode="fine", cache_dir=None):
     """Loads the CIFAR100 dataset.
 
     This is a dataset of 50,000 32x32 color training images and
@@ -39,6 +39,8 @@ def load_data(label_mode="fine"):
       label_mode: one of "fine", "coarse". If it is "fine" the category labels
         are the fine-grained labels, if it is "coarse" the output labels are the
         coarse-grained superclasses.
+      cache_dir: directory where to cache the dataset locally. When None,
+        defaults to `~/.keras/datasets`.
 
     Returns:
       Tuple of NumPy arrays: `(x_train, y_train), (x_test, y_test)`.
@@ -75,6 +77,9 @@ def load_data(label_mode="fine"):
 
     dirname = "cifar-100-python"
     origin = "https://www.cs.toronto.edu/~kriz/cifar-100-python.tar.gz"
+    if cache_dir:
+        cache_dir = os.path.expanduser(cache_dir)
+        os.makedirs(cache_dir, exist_ok=True)
     path = get_file(
         dirname,
         origin=origin,
@@ -82,6 +87,7 @@ def load_data(label_mode="fine"):
         file_hash=(  # noqa: E501
             "85cd44d02ba6437773c5bbd22e183051d648de2e7d6b014e1ef29b855ba677a7"
         ),
+        cache_dir=cache_dir,
     )
 
     fpath = os.path.join(path, "train")

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/datasets/fashion_mnist.py

@@ -26,7 +26,7 @@ from tensorflow.python.util.tf_export import keras_export
 
 
 @keras_export("keras.datasets.fashion_mnist.load_data")
-def load_data():
+def load_data(cache_dir=None):
     """Loads the Fashion-MNIST dataset.
 
     This is a dataset of 60,000 28x28 grayscale images of 10 fashion categories,
@@ -48,6 +48,10 @@ def load_data():
     |   8   | Bag         |
     |   9   | Ankle boot  |
 
+    Args:
+      cache_dir: directory where to cache the dataset locally. When None,
+        defaults to `~/.keras/datasets`.
+
     Returns:
       Tuple of NumPy arrays: `(x_train, y_train), (x_test, y_test)`.
 
@@ -77,7 +81,6 @@ def load_data():
       The copyright for Fashion-MNIST is held by Zalando SE.
       Fashion-MNIST is licensed under the [MIT license](
       https://github.com/zalandoresearch/fashion-mnist/blob/master/LICENSE).
-
     """
     dirname = os.path.join("datasets", "fashion-mnist")
     base = "https://storage.googleapis.com/tensorflow/tf-keras-datasets/"
@@ -87,10 +90,19 @@ def load_data():
         "t10k-labels-idx1-ubyte.gz",
         "t10k-images-idx3-ubyte.gz",
     ]
-
+    if cache_dir:
+        cache_dir = os.path.expanduser(cache_dir)
+        os.makedirs(cache_dir, exist_ok=True)
     paths = []
     for fname in files:
-        paths.append(get_file(fname, origin=base + fname, cache_subdir=dirname))
+        paths.append(
+            get_file(
+                fname,
+                origin=base + fname,
+                cache_dir=cache_dir,
+                cache_subdir=dirname,
+            )
+        )
 
     with gzip.open(paths[0], "rb") as lbpath:
         y_train = np.frombuffer(lbpath.read(), np.uint8, offset=8)

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/datasets/imdb.py

@@ -15,6 +15,7 @@
 """IMDB sentiment classification dataset."""
 
 import json
+import os
 
 import numpy as np
 
@@ -36,6 +37,7 @@ def load_data(
     start_char=1,
     oov_char=2,
     index_from=3,
+    cache_dir=None,
     **kwargs,
 ):
     """Loads the [IMDB dataset](https://ai.stanford.edu/~amaas/data/sentiment/).
@@ -73,6 +75,8 @@ def load_data(
           Words that were cut out because of the `num_words` or
           `skip_top` limits will be replaced with this character.
       index_from: int. Index actual words with this index and higher.
+      cache_dir: directory where to cache the dataset locally. When None,
+          defaults to `~/.keras/datasets`.
       **kwargs: Used for backwards compatibility.
 
     Returns:
@@ -108,12 +112,16 @@ def load_data(
     origin_folder = (
         "https://storage.googleapis.com/tensorflow/tf-keras-datasets/"
     )
+    if cache_dir:
+        cache_dir = os.path.expanduser(cache_dir)
+        os.makedirs(cache_dir, exist_ok=True)
     path = get_file(
         path,
         origin=origin_folder + "imdb.npz",
         file_hash=(  # noqa: E501
             "69664113be75683a8fe16e3ed0ab59fda8886cb3cd7ada244f7d9544e4676b9f"
         ),
+        cache_dir=cache_dir,
     )
     with np.load(path, allow_pickle=True) as f:
         x_train, labels_train = f["x_train"], f["y_train"]

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/datasets/mnist.py

@@ -13,6 +13,7 @@
 # limitations under the License.
 # ==============================================================================
 """MNIST handwritten digits dataset."""
+import os
 
 import numpy as np
 
@@ -23,7 +24,7 @@ from tensorflow.python.util.tf_export import keras_export
 
 
 @keras_export("keras.datasets.mnist.load_data")
-def load_data(path="mnist.npz"):
+def load_data(path="mnist.npz", cache_dir=None):
     """Loads the MNIST dataset.
 
     This is a dataset of 60,000 28x28 grayscale images of the 10 digits,
@@ -32,8 +33,9 @@ def load_data(path="mnist.npz"):
     [MNIST homepage](http://yann.lecun.com/exdb/mnist/).
 
     Args:
-      path: path where to cache the dataset locally
-        (relative to `~/.keras/datasets`).
+      path: path where to cache the dataset locally relative to cache_dir.
+      cache_dir: dir location where to cache the dataset locally. When None,
+        defaults to `~/.keras/datasets`.
 
     Returns:
       Tuple of NumPy arrays: `(x_train, y_train), (x_test, y_test)`.
@@ -72,12 +74,16 @@ def load_data(path="mnist.npz"):
     origin_folder = (
         "https://storage.googleapis.com/tensorflow/tf-keras-datasets/"
     )
+    if cache_dir:
+        cache_dir = os.path.expanduser(cache_dir)
+        os.makedirs(cache_dir, exist_ok=True)
     path = get_file(
         path,
         origin=origin_folder + "mnist.npz",
         file_hash=(  # noqa: E501
             "731c5ac602752760c8e48fbffcf8c3b850d9dc2a2aedcf2cc48468fc17b673d1"
         ),
+        cache_dir=cache_dir,
     )
     with np.load(path, allow_pickle=True) as f:
         x_train, y_train = f["x_train"], f["y_train"]

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/datasets/reuters.py

@@ -15,6 +15,7 @@
 """Reuters topic classification dataset."""
 
 import json
+import os
 
 import numpy as np
 
@@ -37,6 +38,7 @@ def load_data(
     start_char=1,
     oov_char=2,
     index_from=3,
+    cache_dir=None,
     **kwargs,
 ):
     """Loads the Reuters newswire classification dataset.
@@ -83,6 +85,8 @@ def load_data(
           Words that were cut out because of the `num_words` or
           `skip_top` limits will be replaced with this character.
       index_from: int. Index actual words with this index and higher.
+      cache_dir: directory where to cache the dataset locally. When None,
+          defaults to `~/.keras/datasets`.
       **kwargs: Used for backwards compatibility.
 
     Returns:
@@ -114,12 +118,16 @@ def load_data(
     origin_folder = (
         "https://storage.googleapis.com/tensorflow/tf-keras-datasets/"
     )
+    if cache_dir:
+        cache_dir = os.path.expanduser(cache_dir)
+        os.makedirs(cache_dir, exist_ok=True)
     path = get_file(
         path,
         origin=origin_folder + "reuters.npz",
         file_hash=(  # noqa: E501
             "d6586e694ee56d7a4e65172e12b3e987c03096cb01eab99753921ef915959916"
         ),
+        cache_dir=cache_dir,
     )
     with np.load(path, allow_pickle=True) as f:
         xs, labels = f["x"], f["y"]

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/engine/base_layer.py

@@ -578,7 +578,8 @@ class Layer(tf.Module, version_utils.LayerVersionSelector):
             Accepted values are constants defined in the class
             `tf.VariableAggregation`.
           **kwargs: Additional keyword arguments. Accepted values are `getter`,
-            `collections`, `experimental_autocast` and `caching_device`.
+            `collections`, `autocast`, `experimental_autocast` and
+            `caching_device`.
 
         Returns:
           The variable created.
@@ -594,6 +595,7 @@ class Layer(tf.Module, version_utils.LayerVersionSelector):
         # Validate optional keyword arguments.
         for kwarg in kwargs:
             if kwarg not in [
+                "autocast",
                 "collections",
                 "experimental_autocast",
                 "caching_device",
@@ -603,9 +605,13 @@ class Layer(tf.Module, version_utils.LayerVersionSelector):
             ]:
                 raise TypeError("Unknown keyword argument:", kwarg)
         collections_arg = kwargs.pop("collections", None)
-        # 'experimental_autocast' can be set to False by the caller to indicate
-        # an AutoCastVariable should never be created.
-        autocast = kwargs.pop("experimental_autocast", True)
+        # 'autocast' or 'experimental_autocast' can be set to False by the
+        # caller to indicate an AutoCastVariable should never be created.
+        autocast = kwargs.pop("autocast", None)
+        if autocast is None:
+            autocast = kwargs.pop("experimental_autocast", None)
+            if autocast is None:
+                autocast = True
         # See the docstring for tf.Variable about the details for
         # caching_device.
         caching_device = kwargs.pop("caching_device", None)

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/engine/base_layer_v1.py

@@ -352,7 +352,8 @@ class Layer(base_layer.Layer):
             Accepted values are constants defined in the class
             `tf.VariableAggregation`.
           **kwargs: Additional keyword arguments. Accepted values are `getter`,
-            `collections`, `experimental_autocast` and `caching_device`.
+            `collections`, `autocast`, `experimental_autocast` and
+            `caching_device`.
 
         Returns:
           The created variable. Usually either a `Variable` or
@@ -371,6 +372,7 @@ class Layer(base_layer.Layer):
         # Validate optional keyword arguments.
         for kwarg in kwargs:
             if kwarg not in [
+                "autocast",
                 "getter",
                 "collections",
                 "experimental_autocast",
@@ -380,9 +382,13 @@ class Layer(base_layer.Layer):
         has_custom_getter = "getter" in kwargs
         getter = kwargs.pop("getter", base_layer_utils.make_variable)
         collections_arg = kwargs.pop("collections", None)
-        # 'experimental_autocast' can be set to False by the caller to indicate
-        # an AutoCastVariable should never be created.
-        autocast = kwargs.pop("experimental_autocast", True)
+        # 'autocast' or 'experimental_autocast' can be set to False by the
+        # caller to indicate an AutoCastVariable should never be created.
+        autocast = kwargs.pop("autocast", None)
+        if autocast is None:
+            autocast = kwargs.pop("experimental_autocast", None)
+            if autocast is None:
+                autocast = True
         # See the docstring for tf.Variable about the details for
         # caching_device.
         caching_device = kwargs.pop("caching_device", None)

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/engine/node.py

@@ -84,9 +84,10 @@ class Node:
         self.call_args = call_args
         self.call_kwargs = call_kwargs
 
-        # Cached for performance.
+        # Cached for performance. Put kwargs in order of the call method instead
+        # of using the sorted key order from `tf.nest.flatten`.
         self._flat_arguments = tf.nest.flatten(
-            (self.call_args, self.call_kwargs)
+            (self.call_args, self.call_kwargs.values())
         )
         # Used to avoid expensive `nest` operations in the most common case.
         self._single_positional_tensor_passed = (
@@ -176,9 +177,13 @@ class Node:
             for kt_id, kt_index in self._keras_inputs_ids_and_indices:
                 flat_arguments[kt_index] = tensor_dict[kt_id].pop()
 
+            # Pack the same way as `self._flat_arguments`, i.e. `kwargs` as a
+            # list in the original order.
             args, kwargs = tf.nest.pack_sequence_as(
-                (self.call_args, self.call_kwargs), flat_arguments
+                (self.call_args, self.call_kwargs.values()), flat_arguments
             )
+            # Add the keys to `kwargs` to go from a list to a dict.
+            kwargs = {k: v for k, v in zip(self.call_kwargs.keys(), kwargs)}
             return args, kwargs
 
     def serialize(self, make_node_key, node_conversion_map):

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/layers/activation/prelu.py

@@ -102,7 +102,7 @@ class PReLU(Layer):
                 if i not in self.shared_axes:
                     axes[i] = input_shape[i]
         self.input_spec = InputSpec(ndim=len(input_shape), axes=axes)
-        self.built = True
+        super().build(input_shape)
 
     def call(self, inputs):
         pos = backend.relu(inputs)

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/layers/attention/base_dense_attention.py

@@ -86,7 +86,8 @@ class BaseDenseAttention(base_layer.BaseRandomLayer):
         # be purely stateless, with no reference to any variable.
         if self.dropout > 0:
             super().build(input_shape)
-        self.built = True
+        else:
+            base_layer.Layer.build(self, input_shape)
 
     def _calculate_scores(self, query, key):
         """Calculates attention scores.

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/layers/convolutional/base_conv.py

@@ -248,7 +248,7 @@ class Conv(Layer):
         self.input_spec = InputSpec(
             min_ndim=self.rank + 2, axes={channel_axis: input_channel}
         )
-        self.built = True
+        super().build(input_shape)
 
     def convolution_op(self, inputs, kernel):
         if self.padding == "causal":

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/layers/convolutional/base_depthwise_conv.py

@@ -20,6 +20,7 @@ import tensorflow.compat.v2 as tf
 from tf_keras.src import constraints
 from tf_keras.src import initializers
 from tf_keras.src import regularizers
+from tf_keras.src.engine.base_layer import Layer
 from tf_keras.src.engine.input_spec import InputSpec
 from tf_keras.src.layers.convolutional.base_conv import Conv
 
@@ -202,7 +203,8 @@ class DepthwiseConv(Conv):
         self.input_spec = InputSpec(
             min_ndim=self.rank + 2, axes={channel_axis: input_dim}
         )
-        self.built = True
+        # Call Layer.build() to skip Conv.build() which we override here.
+        Layer.build(self, input_shape)
 
     def call(self, inputs):
         raise NotImplementedError

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/layers/convolutional/base_separable_conv.py

@@ -21,6 +21,7 @@ from tf_keras.src import activations
 from tf_keras.src import constraints
 from tf_keras.src import initializers
 from tf_keras.src import regularizers
+from tf_keras.src.engine.base_layer import Layer
 from tf_keras.src.engine.input_spec import InputSpec
 from tf_keras.src.layers.convolutional.base_conv import Conv
 
@@ -203,7 +204,8 @@ class SeparableConv(Conv):
             )
         else:
             self.bias = None
-        self.built = True
+        # Call Layer.build() to skip Conv.build() which we override here.
+        Layer.build(self, input_shape)
 
     def call(self, inputs):
         raise NotImplementedError

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/layers/convolutional/conv1d_transpose.py

@@ -22,6 +22,7 @@ from tf_keras.src import constraints
 from tf_keras.src import initializers
 from tf_keras.src import regularizers
 from tf_keras.src.dtensor import utils
+from tf_keras.src.engine.base_layer import Layer
 from tf_keras.src.engine.input_spec import InputSpec
 from tf_keras.src.layers.convolutional.conv1d import Conv1D
 from tf_keras.src.utils import conv_utils
@@ -214,7 +215,8 @@ class Conv1DTranspose(Conv1D):
             )
         else:
             self.bias = None
-        self.built = True
+        # Call Layer.build() to skip Conv.build() which we override here.
+        Layer.build(self, input_shape)
 
     def call(self, inputs):
         inputs_shape = tf.shape(inputs)

{tf_keras-nightly-2.17.0.dev2024031909 → tf_keras_nightly-2.19.0.dev2025011410}/tf_keras/src/layers/convolutional/conv2d_transpose.py

@@ -23,6 +23,7 @@ from tf_keras.src import constraints
 from tf_keras.src import initializers
 from tf_keras.src import regularizers
 from tf_keras.src.dtensor import utils
+from tf_keras.src.engine.base_layer import Layer
 from tf_keras.src.engine.input_spec import InputSpec
 from tf_keras.src.layers.convolutional.conv2d import Conv2D
 from tf_keras.src.utils import conv_utils
@@ -240,7 +241,8 @@ class Conv2DTranspose(Conv2D):
             )
         else:
             self.bias = None
-        self.built = True
+        # Call Layer.build() to skip Conv.build() which we override here.
+        Layer.build(self, input_shape)
 
     def call(self, inputs):
         inputs_shape = tf.shape(inputs)