Update wide_deep_cross_networks example with tensorflow 2.5 changes (k…

…eras-team#456)
satpalsr · May 7, 2021 · 3551725 · 3551725
1 parent d2fb727
commit 3551725
Show file tree

Hide file tree

Showing 3 changed files with 247 additions and 177 deletions.
diff --git a/examples/structured_data/ipynb/wide_deep_cross_networks.ipynb b/examples/structured_data/ipynb/wide_deep_cross_networks.ipynb
@@ -10,7 +10,7 @@
     "\n",
     "**Author:** [Khalid Salama](https://www.linkedin.com/in/khalid-salama-24403144/)<br>\n",
     "**Date created:** 2020/12/31<br>\n",
-    "**Last modified:** 2020/12/31<br>\n",
+    "**Last modified:** 2021/05/05<br>\n",
     "**Description:** Using Wide & Deep and Deep & Cross networks for structured data classification."
    ]
   },
@@ -28,7 +28,7 @@
     "1. [Wide & Deep](https://ai.googleblog.com/2016/06/wide-deep-learning-better-together-with.html) models\n",
     "2. [Deep & Cross](https://arxiv.org/abs/1708.05123) models\n",
     "\n",
-    "Note that this example should be run with TensorFlow 2.3 or higher."
+    "Note that this example should be run with TensorFlow 2.5 or higher."
    ]
   },
   {
@@ -419,7 +419,6 @@
    "outputs": [],
    "source": [
     "\n",
-    "from tensorflow.keras.layers.experimental.preprocessing import CategoryEncoding\n",
     "from tensorflow.keras.layers.experimental.preprocessing import StringLookup\n",
     "\n",
     "\n",
@@ -431,25 +430,25 @@
     "            # Create a lookup to convert string values to an integer indices.\n",
     "            # Since we are not using a mask token nor expecting any out of vocabulary\n",
     "            # (oov) token, we set mask_token to None and  num_oov_indices to 0.\n",
-    "            index = StringLookup(\n",
-    "                vocabulary=vocabulary, mask_token=None, num_oov_indices=0\n",
+    "            lookup = StringLookup(\n",
+    "                vocabulary=vocabulary,\n",
+    "                mask_token=None,\n",
+    "                num_oov_indices=0,\n",
+    "                output_mode=\"int\" if use_embedding else \"binary\",\n",
     "            )\n",
-    "            # Convert the string input values into integer indices.\n",
-    "            value_index = index(inputs[feature_name])\n",
     "            if use_embedding:\n",
+    "                # Convert the string input values into integer indices.\n",
+    "                encoded_feature = lookup(inputs[feature_name])\n",
     "                embedding_dims = int(math.sqrt(len(vocabulary)))\n",
     "                # Create an embedding layer with the specified dimensions.\n",
-    "                embedding_ecoder = layers.Embedding(\n",
+    "                embedding = layers.Embedding(\n",
     "                    input_dim=len(vocabulary), output_dim=embedding_dims\n",
     "                )\n",
     "                # Convert the index values to embedding representations.\n",
-    "                encoded_feature = embedding_ecoder(value_index)\n",
+    "                encoded_feature = embedding(encoded_feature)\n",
     "            else:\n",
-    "                # Create a one-hot encoder.\n",
-    "                onehot_encoder = CategoryEncoding(output_mode=\"binary\")\n",
-    "                onehot_encoder.adapt(index(vocabulary))\n",
-    "                # Convert the index values to a one-hot representation.\n",
-    "                encoded_feature = onehot_encoder(value_index)\n",
+    "                # Convert the string input values into a one hot encoding.\n",
+    "                encoded_feature = lookup(tf.expand_dims(inputs[feature_name], -1))\n",
     "        else:\n",
     "            # Use the numerical features as-is.\n",
     "            encoded_feature = tf.expand_dims(inputs[feature_name], -1)\n",