{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "780b8bd2-ddee-4f38-8507-48643ff1fb3d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.13.0\n"
     ]
    }
   ],
   "source": [
    "import tensorflow\n",
    "print(tensorflow.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "edea97c0-8733-45a0-838e-07bdde3e09c6",
   "metadata": {},
   "outputs": [],
   "source": [
    "mnist = tensorflow.keras.datasets.mnist\n",
    "\n",
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()\n",
    "x_train, x_test = x_train / 255.0, x_test / 255.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "e4f05156-c521-4706-84c5-c759de4243e3",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-28 13:30:52.410401: I tensorflow/core/common_runtime/process_util.cc:146] Creating new thread pool with default inter op setting: 2. Tune using inter_op_parallelism_threads for best performance.\n"
     ]
    }
   ],
   "source": [
    "model = tensorflow.keras.models.Sequential([\n",
    "  tensorflow.keras.layers.Flatten(input_shape=(28, 28)),\n",
    "  tensorflow.keras.layers.Dense(128, activation='relu'),\n",
    "  tensorflow.keras.layers.Dropout(0.2),\n",
    "  tensorflow.keras.layers.Dense(10)\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "0af8cfcd-1025-45cc-83bd-1af251495622",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 0.51568913, -0.22694974,  0.7567456 , -0.8453866 ,  0.40375337,\n",
       "         0.06754087,  1.095877  , -0.82267034,  0.59129924,  0.15186527]],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions = model(x_train[:1]).numpy()\n",
    "predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "afab32b7-ef89-4f5b-8641-4d15f31137d3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.571601"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "loss_fn = tensorflow.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n",
    "loss_fn(y_train[:1], predictions).numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "e9c2c641-62ed-483b-a60a-73b7bef90052",
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(\n",
    "    optimizer='adam',\n",
    "    loss=loss_fn,\n",
    "    metrics=['accuracy']\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "6cc7246e-e875-4d84-9447-2f4349e0d1a3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start time 2024-04-28 13:30:52.532594\n",
      "Epoch 1/5\n",
      "1875/1875 [==============================] - 10s 5ms/step - loss: 0.2980 - accuracy: 0.9135\n",
      "Epoch 2/5\n",
      "1875/1875 [==============================] - 33s 17ms/step - loss: 0.1455 - accuracy: 0.9574\n",
      "Epoch 3/5\n",
      "1875/1875 [==============================] - 19s 10ms/step - loss: 0.1104 - accuracy: 0.9668\n",
      "Epoch 4/5\n",
      "1875/1875 [==============================] - 17s 9ms/step - loss: 0.0874 - accuracy: 0.9724\n",
      "Epoch 5/5\n",
      "1875/1875 [==============================] - 17s 9ms/step - loss: 0.0774 - accuracy: 0.9759\n",
      "End time 2024-04-28 13:33:15.639799\n",
      "0:02:23.107205\n"
     ]
    }
   ],
   "source": [
    "import datetime\n",
    "\n",
    "start = datetime.datetime.now()\n",
    "print(f\"Start time {start}\")\n",
    "\n",
    "model.fit(x_train, y_train, epochs=5)\n",
    "\n",
    "end = datetime.datetime.now()\n",
    "print(f\"End time {end}\")\n",
    "print(end - start)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "1db98023-3052-4f26-8af0-4725eae0fad6",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-28 13:40:14.388937: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
      "To enable the following instructions: AVX2 AVX512F FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2024-04-28 13:41:07.064024: I tensorflow/core/common_runtime/process_util.cc:146] Creating new thread pool with default inter op setting: 2. Tune using inter_op_parallelism_threads for best performance.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start time 2024-04-28 13:41:08.409854\n",
      "End time 2024-04-28 13:42:22.532379\n",
      "0:01:14.122525\n"
     ]
    }
   ],
   "source": [
    "import tensorflow\n",
    "import datetime\n",
    "\n",
    "mnist = tensorflow.keras.datasets.mnist\n",
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()\n",
    "x_train, x_test = x_train / 255.0, x_test / 255.0\n",
    "\n",
    "model = tensorflow.keras.models.Sequential([\n",
    "  tensorflow.keras.layers.Flatten(input_shape=(28, 28)),\n",
    "  tensorflow.keras.layers.Dense(128, activation='relu'),\n",
    "  tensorflow.keras.layers.Dropout(0.2),\n",
    "  tensorflow.keras.layers.Dense(10)\n",
    "])\n",
    "\n",
    "loss_fn = tensorflow.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n",
    "\n",
    "optimizer = tensorflow.keras.optimizers.Adam()\n",
    "\n",
    "@tensorflow.function\n",
    "def train_step(x, y):\n",
    "    with tensorflow.GradientTape() as tape:\n",
    "        predictions = model(x, training=True)\n",
    "        loss = loss_fn(y, predictions)\n",
    "    gradients = tape.gradient(loss, model.trainable_variables)\n",
    "    optimizer.apply_gradients(zip(gradients, model.trainable_variables))\n",
    "\n",
    "batch_size = 32\n",
    "dataset = tensorflow.data.Dataset.from_tensor_slices((x_train, y_train)).batch(batch_size)\n",
    "\n",
    "start = datetime.datetime.now()\n",
    "print(f\"Start time {start}\")\n",
    "\n",
    "epochs = 5\n",
    "for epoch in range(epochs):\n",
    "    for x_batch, y_batch in dataset:\n",
    "        train_step(x_batch, y_batch)\n",
    "\n",
    "end = datetime.datetime.now()\n",
    "print(f\"End time {end}\")\n",
    "print(end - start)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "62988706-fd36-4cf5-b80a-b4adf3ee3653",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fca4de9a-798c-46e9-bf70-b51cb7c79ab8",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Tensorflow (Intel® oneAPI 2023.2)",
   "language": "python",
   "name": "c009-intel_distribution_of_python_3_oneapi-beta05-tf"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}