Update TensorFlow Privacy to use Python 3 super().

PiperOrigin-RevId: 424916118

tensorflow_privacy/privacy/analysis/rdp_privacy_accountant.py

@@ -527,7 +527,7 @@ class RdpAccountant(privacy_accountant.PrivacyAccountant):
       orders: Optional[Collection[float]] = None,
       neighboring_relation: NeighborRel = NeighborRel.ADD_OR_REMOVE_ONE,
   ):
-    super(RdpAccountant, self).__init__(neighboring_relation)
+    super().__init__(neighboring_relation)
     if orders is None:
       # Default orders chosen to give good coverage for Gaussian mechanism in
       # the privacy regime of interest. In the future, more orders might be

tensorflow_privacy/privacy/bolt_on/losses.py

@@ -138,10 +138,7 @@ class StrongConvexHuber(losses.Loss, StrongConvexMixin):
     self.radius_constant = radius_constant
     self.dtype = dtype
     self.reg_lambda = tf.constant(reg_lambda, dtype=self.dtype)
-    super(StrongConvexHuber, self).__init__(
+    super().__init__(reduction=reduction, name="strongconvexhuber")
-        name="strongconvexhuber",
-        reduction=reduction,
-    )
 
   def call(self, y_true, y_pred):
     """Computes loss.
@@ -248,7 +245,7 @@ class StrongConvexBinaryCrossentropy(
     self.dtype = dtype
     self.C = c_arg  # pylint: disable=invalid-name
     self.reg_lambda = tf.constant(reg_lambda, dtype=self.dtype)
-    super(StrongConvexBinaryCrossentropy, self).__init__(
+    super().__init__(
         reduction=reduction,
         name="strongconvexbinarycrossentropy",
         from_logits=from_logits,

tensorflow_privacy/privacy/bolt_on/models_test.py

@@ -28,7 +28,7 @@ class TestLoss(losses.Loss, StrongConvexMixin):
   """Test loss function for testing BoltOn model."""
 
   def __init__(self, reg_lambda, c_arg, radius_constant, name='test'):
-    super(TestLoss, self).__init__(name=name)
+    super().__init__(name=name)
     self.reg_lambda = reg_lambda
     self.C = c_arg  # pylint: disable=invalid-name
     self.radius_constant = radius_constant
@@ -103,7 +103,7 @@ class TestOptimizer(OptimizerV2):
   """Test optimizer used for testing BoltOn model."""
 
   def __init__(self):
-    super(TestOptimizer, self).__init__('test')
+    super().__init__('test')
 
   def compute_gradients(self):
     return 0

tensorflow_privacy/privacy/bolt_on/optimizers_test.py

@@ -51,7 +51,7 @@ class TestModel(Model):  # pylint: disable=abstract-method
       input_shape:
       init_value:
     """
-    super(TestModel, self).__init__(name='bolton', dynamic=False)
+    super().__init__(name='bolton', dynamic=False)
     self.n_outputs = n_outputs
     self.layer_input_shape = input_shape
     self.output_layer = tf.keras.layers.Dense(
@@ -142,7 +142,7 @@ class TestOptimizer(OptimizerV2):
   """Optimizer used for testing the BoltOn optimizer."""
 
   def __init__(self):
-    super(TestOptimizer, self).__init__('test')
+    super().__init__('test')
     self.not_private = 'test'
     self.iterations = tf.constant(1, dtype=tf.float32)
     self._iterations = tf.constant(1, dtype=tf.float32)

tensorflow_privacy/privacy/dp_query/nested_query.py

@@ -138,4 +138,4 @@ class NestedSumQuery(NestedQuery, dp_query.SumAggregationDPQuery):
 
     tree.map_structure(check, queries)
 
-    super(NestedSumQuery, self).__init__(queries)
+    super().__init__(queries)

tensorflow_privacy/privacy/dp_query/no_privacy_query.py

@@ -45,8 +45,7 @@ class NoPrivacyAverageQuery(dp_query.SumAggregationDPQuery):
 
   def initial_sample_state(self, template):
     """Implements `tensorflow_privacy.DPQuery.initial_sample_state`."""
-    return (super(NoPrivacyAverageQuery,
+    return super().initial_sample_state(template), tf.constant(0.0)
-                  self).initial_sample_state(template), tf.constant(0.0))
 
   def preprocess_record(self, params, record, weight=1):
     """Implements `tensorflow_privacy.DPQuery.preprocess_record`.

tensorflow_privacy/privacy/dp_query/quantile_estimator_query.py

@@ -198,7 +198,7 @@ class NoPrivacyQuantileEstimatorQuery(QuantileEstimatorQuery):
         updating is preferred for non-negative records like vector norms that
         could potentially be very large or very close to zero.
     """
-    super(NoPrivacyQuantileEstimatorQuery, self).__init__(
+    super().__init__(
         initial_estimate,
         target_quantile,
         learning_rate,

tensorflow_privacy/privacy/estimators/binary_class_head.py

@@ -32,7 +32,7 @@ class DPBinaryClassHead(tf.estimator.BinaryClassHead):
                loss_reduction=tf.keras.losses.Reduction.SUM_OVER_BATCH_SIZE,
                loss_fn=None,
                name=None):
-    super(DPBinaryClassHead, self).__init__(
+    super().__init__(
         weight_column=weight_column,
         thresholds=thresholds,
         label_vocabulary=label_vocabulary,

tensorflow_privacy/privacy/estimators/dnn.py

@@ -61,7 +61,7 @@ class DNNClassifier(tf.estimator.Estimator):
           config=config,
           batch_norm=batch_norm)
 
-    super(DNNClassifier, self).__init__(
+    super().__init__(
         model_fn=_model_fn,
         model_dir=model_dir,
         config=config,

tensorflow_privacy/privacy/estimators/multi_class_head.py

@@ -32,7 +32,7 @@ class DPMultiClassHead(tf.estimator.MultiClassHead):
                loss_reduction=losses_utils.ReductionV2.SUM_OVER_BATCH_SIZE,
                loss_fn=None,
                name=None):
-    super(DPMultiClassHead, self).__init__(
+    super().__init__(
         n_classes=n_classes,
         weight_column=weight_column,
         label_vocabulary=label_vocabulary,

tensorflow_privacy/privacy/estimators/multi_label_head.py

@@ -36,7 +36,7 @@ class DPMultiLabelHead(tf.estimator.MultiLabelHead):
                name=None):
     if loss_reduction == tf.keras.losses.Reduction.NONE:
       loss_reduction = tf.keras.losses.Reduction.SUM_OVER_BATCH_SIZE
-    super(DPMultiLabelHead, self).__init__(
+    super().__init__(
         n_classes=n_classes,
         weight_column=weight_column,
         thresholds=thresholds,

tensorflow_privacy/privacy/estimators/v1/dnn.py

@@ -62,7 +62,7 @@ class DNNClassifier(tf.estimator.Estimator):
           config=config,
           batch_norm=batch_norm)
 
-    super(DNNClassifier, self).__init__(
+    super().__init__(
         model_fn=_model_fn,
         model_dir=model_dir,
         config=config,

tensorflow_privacy/privacy/keras_models/dp_keras_model.py

@@ -76,7 +76,7 @@ def make_dp_model_class(cls):
             **kwargs: These will be passed on to the base class `__init__`
               method.
       """
-      super(DPModelClass, self).__init__(*args, **kwargs)
+      super().__init__(*args, **kwargs)
       self._l2_norm_clip = l2_norm_clip
       self._noise_multiplier = noise_multiplier
 

tensorflow_privacy/privacy/optimizers/dp_optimizer.py

@@ -115,7 +115,7 @@ def make_optimizer_class(cls):
         *args: These will be passed on to the base class `__init__` method.
         **kwargs: These will be passed on to the base class `__init__` method.
       """
-      super(DPOptimizerClass, self).__init__(*args, **kwargs)
+      super().__init__(*args, **kwargs)
       self._dp_sum_query = dp_sum_query
       self._num_microbatches = num_microbatches
       self._global_state = None

tensorflow_privacy/privacy/optimizers/dp_optimizer_eager_test.py

@@ -24,7 +24,7 @@ class DPOptimizerEagerTest(tf.test.TestCase, parameterized.TestCase):
 
   def setUp(self):
     tf.enable_eager_execution()
-    super(DPOptimizerEagerTest, self).setUp()
+    super().setUp()
 
   def _loss_fn(self, val0, val1):
     return 0.5 * tf.reduce_sum(

tensorflow_privacy/privacy/optimizers/dp_optimizer_keras.py

@@ -151,7 +151,7 @@ def make_keras_optimizer_class(cls):
         *args: These will be passed on to the base class `__init__` method.
         **kwargs: These will be passed on to the base class `__init__` method.
       """
-      super(DPOptimizerClass, self).__init__(*args, **kwargs)
+      super().__init__(*args, **kwargs)
       self.gradient_accumulation_steps = gradient_accumulation_steps
       self._l2_norm_clip = l2_norm_clip
       self._noise_multiplier = noise_multiplier
@@ -162,14 +162,13 @@ def make_keras_optimizer_class(cls):
       self._was_dp_gradients_called = False
 
     def _create_slots(self, var_list):
-      super(DPOptimizerClass, self)._create_slots(var_list)
+      super()._create_slots(var_list)
       if self.gradient_accumulation_steps > 1:
         for var in var_list:
           self.add_slot(var, 'grad_acc')
 
     def _prepare_local(self, var_device, var_dtype, apply_state):
-      super(DPOptimizerClass, self)._prepare_local(
+      super()._prepare_local(var_device, var_dtype, apply_state)
-          var_device, var_dtype, apply_state)
       if self.gradient_accumulation_steps > 1:
         apply_update = tf.math.equal(
             tf.math.floormod(self.iterations + 1,

tensorflow_privacy/privacy/optimizers/dp_optimizer_keras_vectorized.py

@@ -117,7 +117,7 @@ def make_vectorized_keras_optimizer_class(cls):
         *args: These will be passed on to the base class `__init__` method.
         **kwargs: These will be passed on to the base class `__init__` method.
       """
-      super(DPOptimizerClass, self).__init__(*args, **kwargs)
+      super().__init__(*args, **kwargs)
       self._l2_norm_clip = l2_norm_clip
       self._noise_multiplier = noise_multiplier
       self._num_microbatches = num_microbatches
@@ -204,7 +204,7 @@ def make_vectorized_keras_optimizer_class(cls):
       def process_microbatch(microbatch_loss):
         """Compute clipped grads for one microbatch."""
         mean_loss = tf.reduce_mean(input_tensor=microbatch_loss)
-        grads = super(DPOptimizerClass, self).get_gradients(mean_loss, params)
+        grads = cls.get_gradients(self, mean_loss, params)
         grads_list = [
             g if g is not None else tf.zeros_like(v)
             for (g, v) in zip(list(grads), params)

tensorflow_privacy/privacy/optimizers/dp_optimizer_vectorized.py

@@ -100,7 +100,7 @@ def make_vectorized_optimizer_class(cls):
         *args: These will be passed on to the base class `__init__` method.
         **kwargs: These will be passed on to the base class `__init__` method.
       """
-      super(DPOptimizerClass, self).__init__(*args, **kwargs)
+      super().__init__(*args, **kwargs)
       self._l2_norm_clip = l2_norm_clip
       self._noise_multiplier = noise_multiplier
       self._num_microbatches = num_microbatches

tensorflow_privacy/privacy/privacy_tests/membership_inference_attack/codelabs/third_party/seq2seq_membership_inference/seq2seq_membership_inference_codelab.ipynb

@@ -404,7 +404,7 @@
         "                 **kwargs):\n",
         "        if 'input_shape' not in kwargs and 'input_dim' in kwargs:\n",
         "            kwargs['input_shape'] = (kwargs.pop('input_dim'),)\n",
-        "        super(Attention, self).__init__(**kwargs)\n",
+        "        super().__init__(**kwargs)\n",
         "        self.units = units\n",
         "        self.activation = activations.get(activation)\n",
         "        self.use_bias = use_bias\n",
@@ -523,7 +523,7 @@
         "            'kernel_constraint': constraints.serialize(self.kernel_constraint),\n",
         "            'bias_constraint': constraints.serialize(self.bias_constraint)\n",
         "        }\n",
-        "        base_config = super(Attention, self).get_config()\n",
+        "        base_config = super().get_config()\n",
         "        return dict(list(base_config.items()) + list(config.items()))"
       ]
     },
@@ -553,7 +553,7 @@
         "                 **kwargs):\n",
         "        if 'input_shape' not in kwargs and 'input_dim' in kwargs:\n",
         "            kwargs['input_shape'] = (kwargs.pop('input_dim'),)\n",
-        "        super(DenseTransposeTied, self).__init__(**kwargs)\n",
+        "        super().__init__(**kwargs)\n",
         "        self.units = units\n",
         "        # We add these two properties to save the tied weights\n",
         "        self.tied_to = tied_to\n",

tensorflow_privacy/privacy/privacy_tests/membership_inference_attack/data_structures_test.py

@@ -244,7 +244,7 @@ class SingleMembershipProbabilityResultTest(absltest.TestCase):
 class AttackResultsCollectionTest(absltest.TestCase):
 
   def __init__(self, *args, **kwargs):
-    super(AttackResultsCollectionTest, self).__init__(*args, **kwargs)
+    super().__init__(*args, **kwargs)
 
     self.some_attack_result = SingleAttackResult(
         slice_spec=SingleSliceSpec(None),

tensorflow_privacy/privacy/privacy_tests/membership_inference_attack/privacy_report_test.py

@@ -29,7 +29,7 @@ from tensorflow_privacy.privacy.privacy_tests.membership_inference_attack.data_s
 class PrivacyReportTest(absltest.TestCase):
 
   def __init__(self, *args, **kwargs):
-    super(PrivacyReportTest, self).__init__(*args, **kwargs)
+    super().__init__(*args, **kwargs)
 
     # Classifier that achieves an AUC of 0.5.
     self.imperfect_classifier_result = SingleAttackResult(

tutorials/bolton_tutorial.py

@@ -144,7 +144,7 @@ bolt.fit_generator(generator,
 class TestModel(tf.keras.Model):  # pylint: disable=abstract-method
 
   def __init__(self, reg_layer, number_of_outputs=1):
-    super(TestModel, self).__init__(name='test')
+    super().__init__(name='test')
     self.output_layer = tf.keras.layers.Dense(number_of_outputs,
                                               kernel_regularizer=reg_layer)