replace tf_keras with tf.keras

Author: othakkar

AI-and-Analytics/Features-and-Functionality/IntelTensorFlow_Enabling_Auto_Mixed_Precision_for_TransferLearning/enabling_automixed_precision_for_transfer_learning_with_tensorflow.ipynb

@@ -32,7 +32,6 @@
     "import tensorflow_hub as hub\n",
     "from datetime import datetime\n",
     "import requests\n",
-    "import tf_keras\n",
     "print(\"We are using Tensorflow version: \", tf.__version__)"
    ]
   },
@@ -99,7 +98,7 @@
    "outputs": [],
    "source": [
     "tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.ERROR)\n",
-    "data_root = tf_keras.utils.get_file(\n",
+    "data_root = tf.keras.utils.get_file(\n",
     "  'flower_photos',\n",
     "  'https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz',\n",
     "   untar=True)\n",
@@ -108,7 +107,7 @@
     "img_height = 224\n",
     "img_width = 224\n",
     "\n",
-    "train_ds = tf_keras.utils.image_dataset_from_directory(\n",
+    "train_ds = tf.keras.utils.image_dataset_from_directory(\n",
     "  str(data_root),\n",
     "  validation_split=0.2,\n",
     "  subset=\"training\",\n",
@@ -117,7 +116,7 @@
     "  batch_size=batch_size\n",
     ")\n",
     "\n",
-    "val_ds = tf_keras.utils.image_dataset_from_directory(\n",
+    "val_ds = tf.keras.utils.image_dataset_from_directory(\n",
     "  str(data_root),\n",
     "  validation_split=0.2,\n",
     "  subset=\"validation\",\n",
@@ -147,7 +146,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "normalization_layer = tf_keras.layers.Rescaling(1./255)\n",
+    "normalization_layer = tf.keras.layers.Rescaling(1./255)\n",
     "train_ds = train_ds.map(lambda x, y: (normalization_layer(x), y)) # Where x—images, y—labels.\n",
     "val_ds = val_ds.map(lambda x, y: (normalization_layer(x), y)) # Where x—images, y—labels.\n",
     "\n",
@@ -221,7 +220,7 @@
    "id": "70b3eb9b",
    "metadata": {},
    "source": [
-    "Attach the last fully connected classification layer in a **tf_keras.Sequential** model."
+    "Attach the last fully connected classification layer in a **tf.keras.Sequential** model."
    ]
   },
   {
@@ -233,14 +232,14 @@
    "source": [
     "num_classes = len(class_names)\n",
     "\n",
-    "fp32_model = tf_keras.Sequential([\n",
+    "fp32_model = tf.keras.Sequential([\n",
     "  feature_extractor_layer,\n",
-    "  tf_keras.layers.Dense(num_classes)\n",
+    "  tf.keras.layers.Dense(num_classes)\n",
     "])\n",
     "\n",
     "if arch == 'SPR':\n",
     "    # Create a deep copy of the model to train the bf16 model separately to compare accuracy\n",
-    "    bf16_model = tf_keras.models.clone_model(fp32_model)\n",
+    "    bf16_model = tf.keras.models.clone_model(fp32_model)\n",
     "\n",
     "fp32_model.summary()"
    ]
@@ -260,7 +259,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "class TimeHistory(tf_keras.callbacks.Callback):\n",
+    "class TimeHistory(tf.keras.callbacks.Callback):\n",
     "    def on_train_begin(self, logs={}):\n",
     "        self.times = []\n",
     "        self.throughput = []\n",
@@ -290,8 +289,8 @@
    "outputs": [],
    "source": [
     "fp32_model.compile(\n",
-    "  optimizer=tf_keras.optimizers.SGD(),\n",
-    "  loss=tf_keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
+    "  optimizer=tf.keras.optimizers.SGD(),\n",
+    "  loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
     "  metrics=['acc'])"
    ]
   },
@@ -374,8 +373,8 @@
     "if arch == 'SPR':\n",
     "    # Compile\n",
     "    bf16_model.compile(\n",
-    "      optimizer=tf_keras.optimizers.SGD(),\n",
-    "      loss=tf_keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
+    "      optimizer=tf.keras.optimizers.SGD(),\n",
+    "      loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
     "      metrics=['acc'])\n",
     "    \n",
     "    # Train\n",