Upgrade tensorflow-model-analysis and tensorflow to >=2.16.2.

PiperOrigin-RevId: 705990022

Author: vkarampudi

fairness_indicators/example_model.py

@@ -19,6 +19,8 @@
 results can be visualized using tools like TensorBoard.
 """
 
+from typing import Any
+
 from fairness_indicators import fairness_indicators_metrics  # pylint: disable=unused-import
 from tensorflow import keras
 import tensorflow.compat.v1 as tf
@@ -40,41 +42,58 @@ class ExampleParser(keras.layers.Layer):
 
   def __init__(self, input_feature_key):
     self._input_feature_key = input_feature_key
+    self.input_spec = keras.layers.InputSpec(shape=(1,), dtype=tf.string)
     super().__init__()
 
+  def compute_output_shape(self, input_shape: Any):
+    return [1, 1]
+
   def call(self, serialized_examples):
     def get_feature(serialized_example):
       parsed_example = tf.io.parse_single_example(
           serialized_example, features=FEATURE_MAP
       )
       return parsed_example[self._input_feature_key]
-
+    serialized_examples = tf.cast(serialized_examples, tf.string)
     return tf.map_fn(get_feature, serialized_examples)
 
 
-class ExampleModel(keras.Model):
-  """A Example Keras NLP model."""
+class Reshaper(keras.layers.Layer):
+  """A Keras layer that reshapes the input."""
 
-  def __init__(self, input_feature_key):
-    super().__init__()
-    self.parser = ExampleParser(input_feature_key)
-    self.text_vectorization = keras.layers.TextVectorization(
-        max_tokens=32,
-        output_mode='int',
-        output_sequence_length=32,
-    )
-    self.text_vectorization.adapt(
-        ['nontoxic', 'toxic comment', 'test comment', 'abc', 'abcdef', 'random']
-    )
-    self.dense1 = keras.layers.Dense(32, activation='relu')
-    self.dense2 = keras.layers.Dense(1)
-
-  def call(self, inputs, training=True, mask=None):
-    parsed_example = self.parser(inputs)
-    text_vector = self.text_vectorization(parsed_example)
-    output1 = self.dense1(tf.cast(text_vector, tf.float32))
-    output2 = self.dense2(output1)
-    return output2
+  def call(self, inputs):
+    return tf.reshape(inputs, (1, 32))
+
+
+class Caster(keras.layers.Layer):
+  """A Keras layer that reshapes the input."""
+
+  def call(self, inputs):
+    return tf.cast(inputs, tf.float32)
+
+
+def get_example_model(input_feature_key: str):
+  """Returns a Keras model for testing."""
+  parser = ExampleParser(input_feature_key)
+  text_vectorization = keras.layers.TextVectorization(
+      max_tokens=32,
+      output_mode='int',
+      output_sequence_length=32,
+  )
+  text_vectorization.adapt(
+      ['nontoxic', 'toxic comment', 'test comment', 'abc', 'abcdef', 'random']
+  )
+  dense1 = keras.layers.Dense(32, activation='relu')
+  dense2 = keras.layers.Dense(1)
+
+  inputs = tf.keras.Input(shape=(), dtype=tf.string)
+  parsed_example = parser(inputs)
+  text_vector = text_vectorization(parsed_example)
+  text_vector = Reshaper()(text_vector)
+  text_vector = Caster()(text_vector)
+  output1 = dense1(text_vector)
+  output2 = dense2(output1)
+  return tf.keras.Model(inputs=inputs, outputs=output2)
 
 
 def evaluate_model(

fairness_indicators/example_model_test.py

@@ -81,14 +81,15 @@ def _write_tf_records(self, examples):
 
   def test_example_model(self):
     data = self._create_data()
-    classifier = example_model.ExampleModel(example_model.TEXT_FEATURE)
+    classifier = example_model.get_example_model(example_model.TEXT_FEATURE)
     classifier.compile(optimizer=keras.optimizers.Adam(), loss='mse')
     classifier.fit(
         tf.constant([e.SerializeToString() for e in data]),
         np.array([
             e.features.feature[example_model.LABEL].float_list.value[:][0]
             for e in data
         ]),
+        batch_size=1,
     )
     classifier.save(self._model_dir, save_format='tf')
 

g3doc/tutorials/Fairness_Indicators_TensorBoard_Plugin_Example_Colab.ipynb

@@ -216,7 +216,7 @@
       },
       "outputs": [],
       "source": [
-        "classifier = example_model.ExampleModel(example_model.TEXT_FEATURE)\n",
+        "classifier = example_model.get_example_model(example_model.TEXT_FEATURE)\n",
         "classifier.compile(optimizer=keras.optimizers.Adam(), loss='mse')\n",
         "\n",
         "# Read the data from the training file\n",

setup.py

@@ -38,16 +38,20 @@ def select_constraint(default, nightly=None, git_master=None):
     return default
 
 REQUIRED_PACKAGES = [
-    'tensorflow>=2.15,<2.16',
+    'tensorflow>=2.16.2,<2.17.0',
     'tensorflow-hub>=0.16.1,<1.0.0',
-    'tensorflow-data-validation' + select_constraint(
-        default='>=1.15.1,<2.0.0',
-        nightly='>=1.16.0.dev',
-        git_master='@git+https://github.com/tensorflow/data-validation@master'),
-    'tensorflow-model-analysis' + select_constraint(
-        default='>=0.46,<0.47',
-        nightly='>=0.47.0.dev',
-        git_master='@git+https://github.com/tensorflow/model-analysis@master'),
+    'tensorflow-data-validation'
+    + select_constraint(
+        default='>=1.16.1,<2.0.0',
+        nightly='>=1.17.0.dev',
+        git_master='@git+https://github.com/tensorflow/data-validation@master',
+    ),
+    'tensorflow-model-analysis'
+    + select_constraint(
+        default='>=0.47,<0.48',
+        nightly='>=0.48.0.dev',
+        git_master='@git+https://github.com/tensorflow/model-analysis@master',
+    ),
     'witwidget>=1.4.4,<2',
     'protobuf>=3.20.3,<5',
 ]

tensorboard_plugin/setup.py

@@ -43,12 +43,12 @@ def select_constraint(default, nightly=None, git_master=None):
 
 REQUIRED_PACKAGES = [
     'protobuf>=3.20.3,<5',
-    'tensorboard>=2.15.2,<2.16.0',
-    'tensorflow>=2.15,<2.16',
+    'tensorboard>=2.16.2,<2.17.0',
+    'tensorflow>=2.16.2,<2.17.0',
     'tensorflow-model-analysis'
     + select_constraint(
-        default='>=0.46,<0.47',
-        nightly='>=0.47.0.dev',
+        default='>=0.47,<0.48',
+        nightly='>=0.48.0.dev',
         git_master='@git+https://github.com/tensorflow/model-analysis@master',
     ),
     'werkzeug<2',