Minor update to mnist_lr_tutorial.py to avoid (some) deprecated items.

PiperOrigin-RevId: 339327388

tutorials/mnist_lr_tutorial.py

@@ -30,8 +30,6 @@ import math
 from absl import app
 from absl import flags
 
-from distutils.version import LooseVersion
-
 import numpy as np
 import tensorflow.compat.v1 as tf
 
@@ -39,10 +37,7 @@ from tensorflow_privacy.privacy.analysis.rdp_accountant import compute_rdp
 from tensorflow_privacy.privacy.analysis.rdp_accountant import get_privacy_spent
 from tensorflow_privacy.privacy.optimizers import dp_optimizer
 
-if LooseVersion(tf.__version__) < LooseVersion('2.0.0'):
-  GradientDescentOptimizer = tf.train.GradientDescentOptimizer
-else:
-  GradientDescentOptimizer = tf.optimizers.SGD  # pylint: disable=invalid-name
+GradientDescentOptimizer = tf.train.GradientDescentOptimizer
 
 FLAGS = flags.FLAGS
 
@@ -62,14 +57,11 @@ flags.DEFINE_float('data_l2_norm', 8, 'Bound on the L2 norm of normalized data')
 def lr_model_fn(features, labels, mode, nclasses, dim):
   """Model function for logistic regression."""
   input_layer = tf.reshape(features['x'], tuple([-1]) + dim)
-
-  logits = tf.layers.dense(
-      inputs=input_layer,
+  logits = tf.keras.layers.Dense(
       units=nclasses,
-      kernel_regularizer=tf.contrib.layers.l2_regularizer(
-          scale=FLAGS.regularizer),
-      bias_regularizer=tf.contrib.layers.l2_regularizer(
-          scale=FLAGS.regularizer))
+      kernel_regularizer=tf.keras.regularizers.L2(l2=FLAGS.regularizer),
+      bias_regularizer=tf.keras.regularizers.L2(l2=FLAGS.regularizer)).apply(
+          input_layer)
 
   # Calculate loss as a vector (to support microbatches in DP-SGD).
   vector_loss = tf.nn.sparse_softmax_cross_entropy_with_logits(