add desc

Author: rtb

Diff for: after_optimization.ipynb

@@ -178,6 +178,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Define model\n",
     "model = tf.keras.Sequential([\n",
     "    tf.keras.layers.Conv2D(32, kernel_size=(3,3), input_shape=(1, 28, 28), activation='relu', data_format='channels_first'),\n",
     "    tf.keras.layers.Conv2D(64, kernel_size=(3,3), activation='relu'),\n",
@@ -217,7 +218,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# Define train epochs\n",
+    "# Define epochs\n",
     "epochs = 100\n",
     "\n",
     "# Compile model with XLA\n",
@@ -252,6 +253,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Define model save path\n",
     "model_save_path = r'./latest.h5'\n",
     "\n",
     "# save model to file\n",

Diff for: after_optimization_multi.ipynb

@@ -220,6 +220,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Define and compile model in strategy scope\n",
     "with strategy.scope():\n",
     "    model = tf.keras.Sequential([\n",
     "        tf.keras.layers.Conv2D(32, kernel_size=(3,3), input_shape=(1, 28, 28), activation='relu', data_format='channels_first'),\n",
@@ -233,7 +234,8 @@
     "    optimizer = tf.keras.optimizers.AdamW(learning_rate=0.001)\n",
     "    loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n",
     "    metrics = ['accuracy']\n",
-    "\n",
+    "    \n",
+    "    # Compile model with XLA\n",
     "    model.compile(optimizer=optimizer,\n",
     "                loss=loss_fn,\n",
     "                metrics=metrics,\n",
@@ -260,8 +262,13 @@
    "outputs": [],
    "source": [
     "epochs = 100\n",
+    "\n",
+    "# Compute interation per epoch for 'large but limited size dataset'\n",
+    "# If dataset's size is infinite, set how many step to do on 1 epoch\n",
     "iteration_per_epoch = math.ceil(len(glob.glob(os.path.join(image_dir, '*/*.png')))/batch_size)\n",
     "\n",
+    "\n",
+    "# start training\n",
     "model.fit(dataloader, epochs=epochs, steps_per_epoch=iteration_per_epoch)\n"
    ]
   },
@@ -284,6 +291,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Define model save path\n",
     "model_save_path = r'./latest.h5'\n",
     "\n",
     "# save model to file\n",

Diff for: before_optimization.ipynb

@@ -181,6 +181,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Define model\n",
     "model = tf.keras.Sequential([\n",
     "    tf.keras.layers.Conv2D(32, kernel_size=(3,3), input_shape=(1, 28, 28), activation='relu', data_format='channels_first'),\n",
     "    tf.keras.layers.Conv2D(64, kernel_size=(3,3), activation='relu'),\n",
@@ -214,6 +215,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Define epochs, batch size, dataset path\n",
     "epochs = 100\n",
     "batch_size = 2560\n",
     "images_path = './mnist_png/training'\n",