diff --git a/12_custom_models_and_training_with_tensorflow.ipynb b/12_custom_models_and_training_with_tensorflow.ipynb
index f87cc36..879879d 100644
--- a/12_custom_models_and_training_with_tensorflow.ipynb
+++ b/12_custom_models_and_training_with_tensorflow.ipynb
@@ -1588,13 +1588,20 @@
     "model.fit(X_train_scaled.astype(np.float32), y_train.astype(np.float32), epochs=2)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "**Warning**: In TF 2.2, tf.keras adds an extra first metric in `model.metrics` at position 0 (see [TF issue #38150](https://github.com/tensorflow/tensorflow/issues/38150)). This forces us to use `model.metrics[-1]` rather than `model.metrics[0]` to access the `HuberMetric`."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 124,
    "metadata": {},
    "outputs": [],
    "source": [
-    "model.metrics[0].threshold"
+    "model.metrics[-1].threshold"
    ]
   },
   {
@@ -1713,6 +1720,13 @@
     "model.fit(X_train_scaled.astype(np.float32), y_train.astype(np.float32), epochs=2)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "**Warning**: In TF 2.2, tf.keras adds an extra first metric in `model.metrics` at position 0 (see [TF issue #38150](https://github.com/tensorflow/tensorflow/issues/38150)). This forces us to use `model.metrics[-1]` rather than `model.metrics[0]` to access the `HuberMetric`."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 134,
@@ -1721,7 +1735,7 @@
    },
    "outputs": [],
    "source": [
-    "model.metrics[0].threshold"
+    "model.metrics[-1].threshold"
    ]
   },
   {
@@ -1976,8 +1990,6 @@
     "class ResidualBlock(keras.layers.Layer):\n",
     "    def __init__(self, n_layers, n_neurons, **kwargs):\n",
     "        super().__init__(**kwargs)\n",
-    "        self.n_layers = n_layers                                     # not shown in the book\n",
-    "        self.n_neurons = n_neurons                                   # not shown\n",
     "        self.hidden = [keras.layers.Dense(n_neurons, activation=\"elu\",\n",
     "                                          kernel_initializer=\"he_normal\")\n",
     "                       for _ in range(n_layers)]\n",
@@ -1986,12 +1998,7 @@
     "        Z = inputs\n",
     "        for layer in self.hidden:\n",
     "            Z = layer(Z)\n",
-    "        return inputs + Z\n",
-    "    \n",
-    "    def get_config(self):                                               # not shown\n",
-    "        base_config = super().get_config()                              # not shown\n",
-    "        return {**base_config,                                          # not shown\n",
-    "                \"n_layers\": self.n_layers, \"n_neurons\": self.n_neurons} # not shown"
+    "        return inputs + Z"
    ]
   },
   {
@@ -2003,7 +2010,6 @@
     "class ResidualRegressor(keras.models.Model):\n",
     "    def __init__(self, output_dim, **kwargs):\n",
     "        super().__init__(**kwargs)\n",
-    "        self.output_dim = output_dim                                 # not shown in the book\n",
     "        self.hidden1 = keras.layers.Dense(30, activation=\"elu\",\n",
     "                                          kernel_initializer=\"he_normal\")\n",
     "        self.block1 = ResidualBlock(2, 30)\n",
@@ -2015,12 +2021,7 @@
     "        for _ in range(1 + 3):\n",
     "            Z = self.block1(Z)\n",
     "        Z = self.block2(Z)\n",
-    "        return self.out(Z)\n",
-    "\n",
-    "    def get_config(self):                                            # not shown\n",
-    "        base_config = super().get_config()                           # not shown\n",
-    "        return {**base_config,                                       # not shown\n",
-    "                \"output_dim\": self.output_dim}                       # not shown"
+    "        return self.out(Z)"
    ]
   },
   {