06_logistic_regression.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "f3d9670e-2a2c-4bf1-a315-9796c6789dc9",
   "metadata": {},
   "source": [
    "# Run logistic regression on fraction labels\n",
    "\n",
    "Run the logistic regression as the first model for the project. It will be used to compare with the deep learning model."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "804cd7c3-db80-4fa9-9d89-4f801ad7d1ed",
   "metadata": {},
   "source": [
    "## Import necessary modules"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "cd85d865-3548-45b8-bf17-c6cdb4234763",
   "metadata": {},
   "outputs": [],
   "source": [
    "# python built-in libraries\n",
    "import datetime"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "56327e98-27db-4fcf-aa87-dde8512b87a2",
   "metadata": {},
   "outputs": [],
   "source": [
    "# choose the tensorflow log level\n",
    "import os\n",
    "os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'  # or any {'0', '1', '2'}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "f930e0a7-cbc7-4f7e-8fe6-d50a90e4dcd1",
   "metadata": {},
   "outputs": [],
   "source": [
    "# import tensorflow\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "f020f42a-6eed-4f9c-8e76-26f15f8fff51",
   "metadata": {},
   "outputs": [],
   "source": [
    "# ohter packages\n",
    "import numpy as np\n",
    "\n",
    "import tfhelper.core as tfhelper"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e3bc5985-ee2c-44f0-9f07-68c483675b77",
   "metadata": {},
   "source": [
    "## Global Variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "82f389f0-549e-4dba-94ae-b963afe7b5fe",
   "metadata": {},
   "outputs": [],
   "source": [
    "# sampling\n",
    "FEATURES_FILE = \"./data/tensors/features_continous.npy\"  # path to the file of the feature tensor\n",
    "LABELS_FILE = \"./data/tensors/labels_continous.npy\"  # path to the file of the feature tensor\n",
    "FRACTIONS = (0.8, 0.1, 0.1)  # train, validation, test\n",
    "BATCH_SIZE = 256  # size of the batch\n",
    "BUFFER_SIZE = BATCH_SIZE * 2  # size of the shuffle buffer\n",
    "# training\n",
    "LEARNING_RATE = 0.001  # starting learning rate\n",
    "BETA1 = 0.9  # decay 1\n",
    "BETA2 = 0.999  # decay 2\n",
    "EPOCHS = 200  # number of epochs\n",
    "# saving\n",
    "TIME_STAMP = datetime.datetime.now().strftime(\"%Y%m%d-%H%M%S\")  # time stamp\n",
    "SAVE_LOC = \"./data/models/lr_{}\".format(TIME_STAMP)  # path to the folder to save the model\n",
    "LOG_LOC = \"./data/logs/fit/lr_{}\".format(TIME_STAMP)  # path to the log, if you change this, you also need to change it in the run_tensorboard"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6e2daffc-3f88-42de-be0e-f614c6e017ba",
   "metadata": {
    "tags": []
   },
   "source": [
    "## Obtain data\n",
    "\n",
    "The data is a 209 data point PDF. The label is a one-hot 2 dim vector. `10` means major phase >= threshold, `01` means major phase <= threshold."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "3d9be3eb-7058-4857-8eea-1cbee6674efa",
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_data():\n",
    "    # load\n",
    "    features, labels = np.load(FEATURES_FILE), np.load(LABELS_FILE)\n",
    "    # sort the labels\n",
    "    labels = np.sort(labels, axis=1)[:, ::-1]\n",
    "    # shuffle\n",
    "    n = features.shape[0]\n",
    "    shuffled_idx = np.random.permutation(n)\n",
    "    features, labels = features[shuffled_idx], labels[shuffled_idx]\n",
    "    # split\n",
    "    f0, f1, f2 = FRACTIONS\n",
    "    i, j, k = round(f0 * n), round((f0 + f1) * n), round((f0 + f1 + f2) * n)\n",
    "    train_data = tf.data.Dataset.from_tensor_slices((features[:i], labels[:i])).shuffle(BUFFER_SIZE).batch(BATCH_SIZE)\n",
    "    valid_data = tf.data.Dataset.from_tensor_slices((features[i:j], labels[i:j])).batch(BATCH_SIZE)\n",
    "    test_data = tf.data.Dataset.from_tensor_slices((features[j:k], labels[j:k])).batch(BATCH_SIZE)\n",
    "    return train_data, valid_data, test_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "76111e4f-5ed3-4318-bee1-9595f8359e5e",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data, valid_data, test_data = load_data()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "66341557-8f2e-4edf-876e-015398a118ee",
   "metadata": {},
   "source": [
    "## Create the model\n",
    "\n",
    "We use the logistric regression. It is a single layer with a softmax function."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "2c9ea8fd-4674-47b1-b871-eec08a87b2c4",
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_model():\n",
    "    model = keras.Sequential()\n",
    "    model.add(keras.Input(shape=(209,)))\n",
    "    model.add(layers.Dense(2, activation=\"softmax\"))\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "505c4f8e-4b54-4343-877f-8bd6485d7cf3",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = create_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "33d75cc4-a197-4b09-a2f7-ef735937c775",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense (Dense)                (None, 2)                 420       \n",
      "=================================================================\n",
      "Total params: 420\n",
      "Trainable params: 420\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "668bbf47-4604-4388-bda2-f7d384b332ae",
   "metadata": {},
   "source": [
    "## Choose optimization method"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "541c52ba-2aa5-48d4-b40c-75d0948b3b2e",
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(\n",
    "    # Optimizer\n",
    "    optimizer=keras.optimizers.Adam(LEARNING_RATE, BETA1, BETA2),\n",
    "    # Loss function to minimize\n",
    "    loss=keras.losses.MeanSquaredError(),\n",
    "    # List of metrics to monitor\n",
    "    metrics=[\n",
    "        keras.metrics.CosineSimilarity()\n",
    "    ],\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "786d396f-3334-433d-99d9-d6026ab524d8",
   "metadata": {
    "tags": []
   },
   "source": [
    "## Train the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "a2ac30aa-afc3-42c7-9f7c-0f057f1496d5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# tensor board\n",
    "tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir=LOG_LOC, histogram_freq=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "4d9f039e-f42a-422a-ab33-b42934d07b66",
   "metadata": {},
   "outputs": [],
   "source": [
    "# early stopping to avoid over fitting\n",
    "earlystopping_callback = keras.callbacks.EarlyStopping(\n",
    "    monitor='val_loss',\n",
    "    patience=50, \n",
    "    restore_best_weights=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "7ecfa6ce-8dd6-4ca9-aab9-d5008b5120b0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "344/344 [==============================] - 10s 21ms/step - loss: 0.0312 - cosine_similarity: 0.9631 - val_loss: 0.0291 - val_cosine_similarity: 0.9653\n",
      "Epoch 2/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0291 - cosine_similarity: 0.9657 - val_loss: 0.0285 - val_cosine_similarity: 0.9666\n",
      "Epoch 3/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0287 - cosine_similarity: 0.9662 - val_loss: 0.0285 - val_cosine_similarity: 0.9657\n",
      "Epoch 4/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0285 - cosine_similarity: 0.9664 - val_loss: 0.0281 - val_cosine_similarity: 0.9665\n",
      "Epoch 5/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0283 - cosine_similarity: 0.9666 - val_loss: 0.0280 - val_cosine_similarity: 0.9665\n",
      "Epoch 6/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0282 - cosine_similarity: 0.9667 - val_loss: 0.0281 - val_cosine_similarity: 0.9662\n",
      "Epoch 7/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0281 - cosine_similarity: 0.9668 - val_loss: 0.0278 - val_cosine_similarity: 0.9668\n",
      "Epoch 8/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0281 - cosine_similarity: 0.9668 - val_loss: 0.0277 - val_cosine_similarity: 0.9673\n",
      "Epoch 9/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0280 - cosine_similarity: 0.9670 - val_loss: 0.0279 - val_cosine_similarity: 0.9663\n",
      "Epoch 10/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0279 - cosine_similarity: 0.9670 - val_loss: 0.0278 - val_cosine_similarity: 0.9665\n",
      "Epoch 11/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0279 - cosine_similarity: 0.9671 - val_loss: 0.0276 - val_cosine_similarity: 0.9672\n",
      "Epoch 12/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0277 - val_cosine_similarity: 0.9678\n",
      "Epoch 13/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0276 - val_cosine_similarity: 0.9676\n",
      "Epoch 14/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0279 - val_cosine_similarity: 0.9662\n",
      "Epoch 15/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9670 - val_loss: 0.0279 - val_cosine_similarity: 0.9663\n",
      "Epoch 16/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0276 - val_cosine_similarity: 0.9670\n",
      "Epoch 17/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 18/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9675\n",
      "Epoch 19/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 20/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 21/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0278 - val_cosine_similarity: 0.9665\n",
      "Epoch 22/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0278 - val_cosine_similarity: 0.9663\n",
      "Epoch 23/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0276 - val_cosine_similarity: 0.9671\n",
      "Epoch 24/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 25/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 26/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0277 - val_cosine_similarity: 0.9667\n",
      "Epoch 27/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 28/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 29/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 30/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9675\n",
      "Epoch 31/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0277 - val_cosine_similarity: 0.9666\n",
      "Epoch 32/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 33/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 34/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 35/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9678\n",
      "Epoch 36/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 37/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 38/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 39/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 40/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9670\n",
      "Epoch 41/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0276 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 42/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 43/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 44/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 45/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 46/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9677\n",
      "Epoch 47/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9678\n",
      "Epoch 48/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 49/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0279 - val_cosine_similarity: 0.9661\n",
      "Epoch 50/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9670\n",
      "Epoch 51/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9676\n",
      "Epoch 52/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9677\n",
      "Epoch 53/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9677\n",
      "Epoch 54/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 55/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9677\n",
      "Epoch 56/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0277 - val_cosine_similarity: 0.9666\n",
      "Epoch 57/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9675\n",
      "Epoch 58/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 59/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 60/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 61/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 62/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 63/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9668\n",
      "Epoch 64/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 65/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 66/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0278 - val_cosine_similarity: 0.9663\n",
      "Epoch 67/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 68/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 69/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 70/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9676\n",
      "Epoch 71/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 72/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 73/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9670\n",
      "Epoch 74/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 75/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 76/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9675\n",
      "Epoch 77/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9675\n",
      "Epoch 78/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 79/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 80/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 81/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0276 - val_cosine_similarity: 0.9670\n",
      "Epoch 82/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 83/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 84/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 85/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 86/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 87/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0276 - val_cosine_similarity: 0.9667\n",
      "Epoch 88/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 89/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 90/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 91/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0276 - cosine_similarity: 0.9673 - val_loss: 0.0279 - val_cosine_similarity: 0.9662\n",
      "Epoch 92/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 93/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0279 - val_cosine_similarity: 0.9662\n",
      "Epoch 94/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 95/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 96/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9670\n",
      "Epoch 97/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 98/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9670\n",
      "Epoch 99/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 100/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0278 - val_cosine_similarity: 0.9663\n",
      "Epoch 101/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 102/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9668\n",
      "Epoch 103/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9668\n",
      "Epoch 104/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9676\n",
      "Epoch 105/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0277 - val_cosine_similarity: 0.9665\n",
      "Epoch 106/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 107/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 108/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 109/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 110/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9667\n",
      "Epoch 111/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 112/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 113/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 114/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9668\n",
      "Epoch 115/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0278 - cosine_similarity: 0.9671 - val_loss: 0.0278 - val_cosine_similarity: 0.9664\n",
      "Epoch 116/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 117/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 118/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 119/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 120/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9675\n",
      "Epoch 121/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 122/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 123/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9670\n",
      "Epoch 124/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9675\n",
      "Epoch 125/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0277 - val_cosine_similarity: 0.9666\n",
      "Epoch 126/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0277 - val_cosine_similarity: 0.9666\n",
      "Epoch 127/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9677\n",
      "Epoch 128/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0276 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 129/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9677\n",
      "Epoch 130/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 131/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9668\n",
      "Epoch 132/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 133/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 134/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0278 - val_cosine_similarity: 0.9663\n",
      "Epoch 135/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 136/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 137/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 138/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 139/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0276 - val_cosine_similarity: 0.9667\n",
      "Epoch 140/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0277 - val_cosine_similarity: 0.9667\n",
      "Epoch 141/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 142/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 143/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 144/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9670\n",
      "Epoch 145/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 146/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 147/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 148/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0276 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 149/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 150/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 151/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9676\n",
      "Epoch 152/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9676\n",
      "Epoch 153/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 154/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 155/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9667\n",
      "Epoch 156/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9668\n",
      "Epoch 157/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9676\n",
      "Epoch 158/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9677\n",
      "Epoch 159/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9670\n",
      "Epoch 160/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 161/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9676\n",
      "Epoch 162/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9675\n",
      "Epoch 163/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0276 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 164/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 165/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9676\n",
      "Epoch 166/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 167/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9670\n",
      "Epoch 168/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 169/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 170/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9677\n",
      "Epoch 171/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0276 - cosine_similarity: 0.9673 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 172/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 173/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0278 - val_cosine_similarity: 0.9663\n",
      "Epoch 174/200\n",
      "344/344 [==============================] - 2s 6ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0277 - val_cosine_similarity: 0.9666\n",
      "Epoch 175/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 176/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 177/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0276 - val_cosine_similarity: 0.9670\n",
      "Epoch 178/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 179/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0276 - cosine_similarity: 0.9673 - val_loss: 0.0276 - val_cosine_similarity: 0.9667\n",
      "Epoch 180/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9671 - val_loss: 0.0275 - val_cosine_similarity: 0.9677\n",
      "Epoch 181/200\n",
      "344/344 [==============================] - 2s 6ms/step - loss: 0.0276 - cosine_similarity: 0.9673 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 182/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0280 - val_cosine_similarity: 0.9660\n",
      "Epoch 183/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 184/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 185/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0278 - val_cosine_similarity: 0.9662\n",
      "Epoch 186/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 187/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 188/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9674\n",
      "Epoch 189/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 190/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9669\n",
      "Epoch 191/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 192/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9676\n",
      "Epoch 193/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9672\n",
      "Epoch 194/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9670\n",
      "Epoch 195/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0276 - val_cosine_similarity: 0.9668\n",
      "Epoch 196/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 197/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9671\n",
      "Epoch 198/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9677\n",
      "Epoch 199/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9672 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n",
      "Epoch 200/200\n",
      "344/344 [==============================] - 2s 5ms/step - loss: 0.0277 - cosine_similarity: 0.9673 - val_loss: 0.0275 - val_cosine_similarity: 0.9673\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(\n",
    "    x=train_data,\n",
    "    epochs=EPOCHS,\n",
    "    callbacks=[\n",
    "        tensorboard_callback,\n",
    "        earlystopping_callback\n",
    "    ],\n",
    "    validation_data=valid_data\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "177de451-f03a-495d-a09d-755626fce563",
   "metadata": {},
   "source": [
    "## Evaluate the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "9d0f540d-a99b-405b-8d05-f138f55a95e5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "43/43 [==============================] - 0s 3ms/step - loss: 0.0277 - cosine_similarity: 0.9668\n"
     ]
    }
   ],
   "source": [
    "result = model.evaluate(\n",
    "    x=test_data,\n",
    "    return_dict=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "296385b0-a5b1-43eb-9aa8-59c7e15a036b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loss                 0.0277\n",
      "cosine_similarity    0.9668\n"
     ]
    }
   ],
   "source": [
    "for name, val in result.items():\n",
    "    print(\"{:20s} {:.4f}\".format(name, val))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "570cabab-3653-4151-a2a7-051680c25de8",
   "metadata": {},
   "source": [
    "## Save the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "416ba8a8-9e6d-4d5e-9685-e211f2f759f4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: ./data/models/lr_20210914-153328/assets\n"
     ]
    }
   ],
   "source": [
    "model.save(SAVE_LOC)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "82f86cd8-cb7a-4913-8b67-e398bcdf2e3a",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:st3107-tfgpu] *",
   "language": "python",
   "name": "conda-env-st3107-tfgpu-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}