intro_to_modeling.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "LY0SJUCSY305"
   },
   "source": [
    "#### Copyright 2018 Google LLC."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "8Q1hsKyBZDVu",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "# Licensed under the Apache License, Version 2.0 (the \"License\");\n",
    "# you may not use this file except in compliance with the License.\n",
    "# You may obtain a copy of the License at\n",
    "#\n",
    "# https://www.apache.org/licenses/LICENSE-2.0\n",
    "#\n",
    "# Unless required by applicable law or agreed to in writing, software\n",
    "# distributed under the License is distributed on an \"AS IS\" BASIS,\n",
    "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n",
    "# See the License for the specific language governing permissions and\n",
    "# limitations under the License."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "y5tftaRtUcm7"
   },
   "source": [
    "#Intro to Modeling\n",
    "\n",
    "\n",
    "**Learning Objectives:**\n",
    "* Become familiar with pandas for handling small datasets\n",
    "* Use the tf.Estimator and Feature Column API to experiment with feature transformations\n",
    "* Use visualizations and run experiments to understand the value of feature transformations\n",
    "\n",
    "Please **make a copy** of this Colab notebook before starting this lab. To do so, choose **File**->**Save a copy in Drive**."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "TT_bZ9E0ZWaN"
   },
   "source": [
    "## Setup\n",
    "\n",
    "Let's start by importing our dependencies."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "cellView": "both",
    "colab": {},
    "colab_type": "code",
    "id": "wZ_T2SgDVKUH",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /Users/eric/.conda/envs/mlbase/lib/python3.8/site-packages/tensorflow/python/compat/v2_compat.py:96: disable_resource_variables (from tensorflow.python.ops.variable_scope) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "non-resource variables are not supported in the long term\n"
     ]
    }
   ],
   "source": [
    "%reset -f\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import math\n",
    "import tensorflow.compat.v1 as tf\n",
    "tf.disable_v2_behavior()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Sx4-PWE-VaD_"
   },
   "source": [
    "## Pandas, a helpful data analysis library for in-memory dataset\n",
    "\n",
    "We use a package called [Pandas](http://pandas.pydata.org/) for reading in our data, exploring our data and doing some basic processing. It is really helpful for datasets that fit in memory! And it has some nice integrations, as you will see.\n",
    "\n",
    "First we set up some options to control how items are displayed and the maximum number of rows to show when displaying a table.  Feel free to change this setup to whatever you'd like."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "hKUIMcPCVRqv",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "# Set pandas output display to have one digit for decimal places and limit it to\n",
    "# printing 15 rows.\n",
    "pd.options.display.float_format = '{:.2f}'.format\n",
    "pd.options.display.max_rows = 15"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "H_fTMztUVelY"
   },
   "source": [
    "### Load the dataset with pandas\n",
    "The car data set we will be using in this lab is provided as a comma separated file without a header row.  In order for each column to have a meaningful header name we must provide it.  We get the information about the columns from the [Automobile Data Set](https://archive.ics.uci.edu/ml/datasets/automobile).\n",
    "\n",
    "We will use the features of the car, to try to predict its price.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "cellView": "code",
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "Y38V73EgVYwt",
    "outputId": "7e3f438d-855e-40e8-a5cf-5575441e6e3e",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data set loaded. Num examples:  205\n"
     ]
    }
   ],
   "source": [
    "# Provide the names for the columns since the CSV file with the data does\n",
    "# not have a header row.\n",
    "feature_names = ['symboling', 'normalized-losses', 'make', 'fuel-type',\n",
    "        'aspiration', 'num-doors', 'body-style', 'drive-wheels',\n",
    "        'engine-location', 'wheel-base', 'length', 'width', 'height', 'weight',\n",
    "        'engine-type', 'num-cylinders', 'engine-size', 'fuel-system', 'bore',\n",
    "        'stroke', 'compression-ratio', 'horsepower', 'peak-rpm', 'city-mpg',\n",
    "        'highway-mpg', 'price']\n",
    "\n",
    "\n",
    "# Load in the data from a CSV file that is comma separated.\n",
    "car_data = pd.read_csv('https://storage.googleapis.com/mledu-datasets/cars_data.csv',\n",
    "                        sep=',', names=feature_names, header=None, encoding='latin-1')\n",
    "\n",
    "\n",
    "# We'll then randomize the data, just to be sure not to get any pathological\n",
    "# ordering effects that might harm the performance of Stochastic Gradient\n",
    "# Descent.\n",
    "car_data = car_data.reindex(np.random.permutation(car_data.index))\n",
    "\n",
    "print(\"Data set loaded. Num examples: \", len(car_data))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "lAYg6gWaNniy",
    "outputId": "ac0a1b13-4cb3-48f3-ef8b-90470471086a",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "len(feature_names)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "nAHZBtDlkmGa"
   },
   "source": [
    "This is a really small dataset! Only 205 examples.\n",
    "\n",
    "For simplicity in this codelab, we do not split the data further into training and validation. But you MUST do this on real datasets, or else you will overfit to your single dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "aQ1HxLrOVqZk"
   },
   "source": [
    "## Task 0: Use pandas to explore and prepare the data\n",
    "\n",
    "- Use Pandas to inspect the data and manually curate a list of numeric_feature_names and categorical_feature_names.\n",
    "\n",
    "\n",
    "Useful functions:\n",
    "- `type()` called on any Python object describes the type of the object\n",
    "- `dataframe[4:7]` pulls out rows 4, 5, 6 in a Pandas dataframe\n",
    "- `dataframe[['mycol1', 'mycol2']]` pulls out the two requested columns into a new Pandas dataframe\n",
    "- `dataframe['mycol1']` returns a Pandas series -- not a dataframe!\n",
    "- `dataframe.describe()` prints out statistics for each dataframe column"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "TfeHYeMf7PwQ",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "car_data[4:7]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "cellView": "both",
    "colab": {},
    "colab_type": "code",
    "id": "VsOUrVozoe9u",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "15\n"
     ]
    }
   ],
   "source": [
    "LABEL = 'price'\n",
    "\n",
    "numeric_feature_names = ['symboling','normalized-losses','wheel-base','engine-size','bore','stroke', 'length', 'width', 'height','weight','compression-ratio','horsepower','peak-rpm','city-mpg','highway-mpg']\n",
    "categorical_feature_names = list(set(feature_names) - set(numeric_feature_names) - set([LABEL]))\n",
    "print(len(numeric_feature_names))\n",
    "# The correct solution will pass these assert statements.\n",
    "assert len(numeric_feature_names) == 15\n",
    "assert len(categorical_feature_names) == 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "cellView": "form",
    "colab": {},
    "colab_type": "code",
    "id": "lLFkHgBm8O1Y",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "#@title Solution (to view code, from cell's menu, select Form -> Show Code)\n",
    "numeric_feature_names = ['symboling', 'normalized-losses', 'wheel-base',\n",
    "        'length', 'width', 'height', 'weight', 'engine-size', 'horsepower',\n",
    "        'peak-rpm', 'city-mpg', 'highway-mpg', 'bore', 'stroke',\n",
    "         'compression-ratio']\n",
    "\n",
    "categorical_feature_names = list(set(feature_names) - set(numeric_feature_names) - set([LABEL]))\n",
    "\n",
    "assert len(numeric_feature_names) == 15\n",
    "assert len(categorical_feature_names) == 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 419
    },
    "colab_type": "code",
    "id": "nabeQFGBpDEN",
    "outputId": "1b782719-d623-4c18-bebd-2a6a36c1768a",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>symboling</th>\n",
       "      <th>normalized-losses</th>\n",
       "      <th>wheel-base</th>\n",
       "      <th>length</th>\n",
       "      <th>width</th>\n",
       "      <th>height</th>\n",
       "      <th>weight</th>\n",
       "      <th>engine-size</th>\n",
       "      <th>horsepower</th>\n",
       "      <th>peak-rpm</th>\n",
       "      <th>city-mpg</th>\n",
       "      <th>highway-mpg</th>\n",
       "      <th>bore</th>\n",
       "      <th>stroke</th>\n",
       "      <th>compression-ratio</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>86</th>\n",
       "      <td>1</td>\n",
       "      <td>125</td>\n",
       "      <td>96.30</td>\n",
       "      <td>172.40</td>\n",
       "      <td>65.40</td>\n",
       "      <td>51.60</td>\n",
       "      <td>2405</td>\n",
       "      <td>122</td>\n",
       "      <td>88</td>\n",
       "      <td>5000</td>\n",
       "      <td>25</td>\n",
       "      <td>32</td>\n",
       "      <td>3.35</td>\n",
       "      <td>3.46</td>\n",
       "      <td>8.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>76</th>\n",
       "      <td>2</td>\n",
       "      <td>161</td>\n",
       "      <td>93.70</td>\n",
       "      <td>157.30</td>\n",
       "      <td>64.40</td>\n",
       "      <td>50.80</td>\n",
       "      <td>1918</td>\n",
       "      <td>92</td>\n",
       "      <td>68</td>\n",
       "      <td>5500</td>\n",
       "      <td>37</td>\n",
       "      <td>41</td>\n",
       "      <td>2.97</td>\n",
       "      <td>3.23</td>\n",
       "      <td>9.40</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>1</td>\n",
       "      <td>101</td>\n",
       "      <td>93.70</td>\n",
       "      <td>150.00</td>\n",
       "      <td>64.00</td>\n",
       "      <td>52.60</td>\n",
       "      <td>1956</td>\n",
       "      <td>92</td>\n",
       "      <td>76</td>\n",
       "      <td>6000</td>\n",
       "      <td>30</td>\n",
       "      <td>34</td>\n",
       "      <td>2.91</td>\n",
       "      <td>3.41</td>\n",
       "      <td>9.20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>103</th>\n",
       "      <td>0</td>\n",
       "      <td>108</td>\n",
       "      <td>100.40</td>\n",
       "      <td>184.60</td>\n",
       "      <td>66.50</td>\n",
       "      <td>55.10</td>\n",
       "      <td>3060</td>\n",
       "      <td>181</td>\n",
       "      <td>152</td>\n",
       "      <td>5200</td>\n",
       "      <td>19</td>\n",
       "      <td>25</td>\n",
       "      <td>3.43</td>\n",
       "      <td>3.27</td>\n",
       "      <td>9.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1</td>\n",
       "      <td>158</td>\n",
       "      <td>105.80</td>\n",
       "      <td>192.70</td>\n",
       "      <td>71.40</td>\n",
       "      <td>55.90</td>\n",
       "      <td>3086</td>\n",
       "      <td>131</td>\n",
       "      <td>140</td>\n",
       "      <td>5500</td>\n",
       "      <td>17</td>\n",
       "      <td>20</td>\n",
       "      <td>3.13</td>\n",
       "      <td>3.40</td>\n",
       "      <td>8.30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>91</th>\n",
       "      <td>1</td>\n",
       "      <td>128</td>\n",
       "      <td>94.50</td>\n",
       "      <td>165.30</td>\n",
       "      <td>63.80</td>\n",
       "      <td>54.50</td>\n",
       "      <td>1918</td>\n",
       "      <td>97</td>\n",
       "      <td>69</td>\n",
       "      <td>5200</td>\n",
       "      <td>31</td>\n",
       "      <td>37</td>\n",
       "      <td>3.15</td>\n",
       "      <td>3.29</td>\n",
       "      <td>9.40</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>105</th>\n",
       "      <td>3</td>\n",
       "      <td>194</td>\n",
       "      <td>91.30</td>\n",
       "      <td>170.70</td>\n",
       "      <td>67.90</td>\n",
       "      <td>49.70</td>\n",
       "      <td>3139</td>\n",
       "      <td>181</td>\n",
       "      <td>200</td>\n",
       "      <td>5200</td>\n",
       "      <td>17</td>\n",
       "      <td>23</td>\n",
       "      <td>3.43</td>\n",
       "      <td>3.27</td>\n",
       "      <td>7.80</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89</th>\n",
       "      <td>1</td>\n",
       "      <td>128</td>\n",
       "      <td>94.50</td>\n",
       "      <td>165.30</td>\n",
       "      <td>63.80</td>\n",
       "      <td>54.50</td>\n",
       "      <td>1889</td>\n",
       "      <td>97</td>\n",
       "      <td>69</td>\n",
       "      <td>5200</td>\n",
       "      <td>31</td>\n",
       "      <td>37</td>\n",
       "      <td>3.15</td>\n",
       "      <td>3.29</td>\n",
       "      <td>9.40</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>193</th>\n",
       "      <td>0</td>\n",
       "      <td>?</td>\n",
       "      <td>100.40</td>\n",
       "      <td>183.10</td>\n",
       "      <td>66.90</td>\n",
       "      <td>55.10</td>\n",
       "      <td>2563</td>\n",
       "      <td>109</td>\n",
       "      <td>88</td>\n",
       "      <td>5500</td>\n",
       "      <td>25</td>\n",
       "      <td>31</td>\n",
       "      <td>3.19</td>\n",
       "      <td>3.40</td>\n",
       "      <td>9.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>144</th>\n",
       "      <td>0</td>\n",
       "      <td>102</td>\n",
       "      <td>97.00</td>\n",
       "      <td>172.00</td>\n",
       "      <td>65.40</td>\n",
       "      <td>54.30</td>\n",
       "      <td>2385</td>\n",
       "      <td>108</td>\n",
       "      <td>82</td>\n",
       "      <td>4800</td>\n",
       "      <td>24</td>\n",
       "      <td>25</td>\n",
       "      <td>3.62</td>\n",
       "      <td>2.64</td>\n",
       "      <td>9.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>205 rows × 15 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     symboling normalized-losses  wheel-base  length  width  height  weight  \\\n",
       "86           1               125       96.30  172.40  65.40   51.60    2405   \n",
       "76           2               161       93.70  157.30  64.40   50.80    1918   \n",
       "34           1               101       93.70  150.00  64.00   52.60    1956   \n",
       "103          0               108      100.40  184.60  66.50   55.10    3060   \n",
       "8            1               158      105.80  192.70  71.40   55.90    3086   \n",
       "..         ...               ...         ...     ...    ...     ...     ...   \n",
       "91           1               128       94.50  165.30  63.80   54.50    1918   \n",
       "105          3               194       91.30  170.70  67.90   49.70    3139   \n",
       "89           1               128       94.50  165.30  63.80   54.50    1889   \n",
       "193          0                 ?      100.40  183.10  66.90   55.10    2563   \n",
       "144          0               102       97.00  172.00  65.40   54.30    2385   \n",
       "\n",
       "     engine-size horsepower peak-rpm  city-mpg  highway-mpg  bore stroke  \\\n",
       "86           122         88     5000        25           32  3.35   3.46   \n",
       "76            92         68     5500        37           41  2.97   3.23   \n",
       "34            92         76     6000        30           34  2.91   3.41   \n",
       "103          181        152     5200        19           25  3.43   3.27   \n",
       "8            131        140     5500        17           20  3.13   3.40   \n",
       "..           ...        ...      ...       ...          ...   ...    ...   \n",
       "91            97         69     5200        31           37  3.15   3.29   \n",
       "105          181        200     5200        17           23  3.43   3.27   \n",
       "89            97         69     5200        31           37  3.15   3.29   \n",
       "193          109         88     5500        25           31  3.19   3.40   \n",
       "144          108         82     4800        24           25  3.62   2.64   \n",
       "\n",
       "     compression-ratio  \n",
       "86                8.50  \n",
       "76                9.40  \n",
       "34                9.20  \n",
       "103               9.00  \n",
       "8                 8.30  \n",
       "..                 ...  \n",
       "91                9.40  \n",
       "105               7.80  \n",
       "89                9.40  \n",
       "193               9.00  \n",
       "144               9.00  \n",
       "\n",
       "[205 rows x 15 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Run to inspect numeric features.\n",
    "car_data[numeric_feature_names]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 436
    },
    "colab_type": "code",
    "id": "b1ss9Q7mpiBy",
    "outputId": "0e48044d-fffc-4637-999f-09ecd563a30c",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>body-style</th>\n",
       "      <th>num-doors</th>\n",
       "      <th>fuel-system</th>\n",
       "      <th>drive-wheels</th>\n",
       "      <th>fuel-type</th>\n",
       "      <th>make</th>\n",
       "      <th>engine-location</th>\n",
       "      <th>engine-type</th>\n",
       "      <th>aspiration</th>\n",
       "      <th>num-cylinders</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>86</th>\n",
       "      <td>sedan</td>\n",
       "      <td>four</td>\n",
       "      <td>2bbl</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>mitsubishi</td>\n",
       "      <td>front</td>\n",
       "      <td>ohc</td>\n",
       "      <td>std</td>\n",
       "      <td>four</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>76</th>\n",
       "      <td>hatchback</td>\n",
       "      <td>two</td>\n",
       "      <td>2bbl</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>mitsubishi</td>\n",
       "      <td>front</td>\n",
       "      <td>ohc</td>\n",
       "      <td>std</td>\n",
       "      <td>four</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>hatchback</td>\n",
       "      <td>two</td>\n",
       "      <td>1bbl</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>honda</td>\n",
       "      <td>front</td>\n",
       "      <td>ohc</td>\n",
       "      <td>std</td>\n",
       "      <td>four</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>103</th>\n",
       "      <td>sedan</td>\n",
       "      <td>four</td>\n",
       "      <td>mpfi</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>nissan</td>\n",
       "      <td>front</td>\n",
       "      <td>ohcv</td>\n",
       "      <td>std</td>\n",
       "      <td>six</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>sedan</td>\n",
       "      <td>four</td>\n",
       "      <td>mpfi</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>audi</td>\n",
       "      <td>front</td>\n",
       "      <td>ohc</td>\n",
       "      <td>turbo</td>\n",
       "      <td>five</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>91</th>\n",
       "      <td>sedan</td>\n",
       "      <td>two</td>\n",
       "      <td>2bbl</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>nissan</td>\n",
       "      <td>front</td>\n",
       "      <td>ohc</td>\n",
       "      <td>std</td>\n",
       "      <td>four</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>105</th>\n",
       "      <td>hatchback</td>\n",
       "      <td>two</td>\n",
       "      <td>mpfi</td>\n",
       "      <td>rwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>nissan</td>\n",
       "      <td>front</td>\n",
       "      <td>ohcv</td>\n",
       "      <td>turbo</td>\n",
       "      <td>six</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89</th>\n",
       "      <td>sedan</td>\n",
       "      <td>two</td>\n",
       "      <td>2bbl</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>nissan</td>\n",
       "      <td>front</td>\n",
       "      <td>ohc</td>\n",
       "      <td>std</td>\n",
       "      <td>four</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>193</th>\n",
       "      <td>wagon</td>\n",
       "      <td>four</td>\n",
       "      <td>mpfi</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>volkswagen</td>\n",
       "      <td>front</td>\n",
       "      <td>ohc</td>\n",
       "      <td>std</td>\n",
       "      <td>four</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>144</th>\n",
       "      <td>sedan</td>\n",
       "      <td>four</td>\n",
       "      <td>2bbl</td>\n",
       "      <td>4wd</td>\n",
       "      <td>gas</td>\n",
       "      <td>subaru</td>\n",
       "      <td>front</td>\n",
       "      <td>ohcf</td>\n",
       "      <td>std</td>\n",
       "      <td>four</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>205 rows × 10 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    body-style num-doors fuel-system drive-wheels fuel-type        make  \\\n",
       "86       sedan      four        2bbl          fwd       gas  mitsubishi   \n",
       "76   hatchback       two        2bbl          fwd       gas  mitsubishi   \n",
       "34   hatchback       two        1bbl          fwd       gas       honda   \n",
       "103      sedan      four        mpfi          fwd       gas      nissan   \n",
       "8        sedan      four        mpfi          fwd       gas        audi   \n",
       "..         ...       ...         ...          ...       ...         ...   \n",
       "91       sedan       two        2bbl          fwd       gas      nissan   \n",
       "105  hatchback       two        mpfi          rwd       gas      nissan   \n",
       "89       sedan       two        2bbl          fwd       gas      nissan   \n",
       "193      wagon      four        mpfi          fwd       gas  volkswagen   \n",
       "144      sedan      four        2bbl          4wd       gas      subaru   \n",
       "\n",
       "    engine-location engine-type aspiration num-cylinders  \n",
       "86            front         ohc        std          four  \n",
       "76            front         ohc        std          four  \n",
       "34            front         ohc        std          four  \n",
       "103           front        ohcv        std           six  \n",
       "8             front         ohc      turbo          five  \n",
       "..              ...         ...        ...           ...  \n",
       "91            front         ohc        std          four  \n",
       "105           front        ohcv      turbo           six  \n",
       "89            front         ohc        std          four  \n",
       "193           front         ohc        std          four  \n",
       "144           front        ohcf        std          four  \n",
       "\n",
       "[205 rows x 10 columns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Run to inspect categorical features.\n",
    "car_data[categorical_feature_names]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "6OjDegBgqNnu",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "# Coerce the numeric features to numbers. This is necessary because the model\n",
    "# crashes because not all the values are numeric.\n",
    "for feature_name in numeric_feature_names + [LABEL]:\n",
    "  car_data[feature_name] = pd.to_numeric(car_data[feature_name], errors='coerce')\n",
    "\n",
    "# Fill missing values with 0.\n",
    "# Is this an OK thing to do? You may want to come back and revisit this decision later.\n",
    "car_data.fillna(0, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Uq-t-8GPvnCW"
   },
   "source": [
    "## Task 1: Make your best model with numeric features. No normalization allowed.\n",
    "\n",
    "Modify the model provided below to achieve the lowest eval loss. You may want to change various hyperparameters:\n",
    "- learning rate\n",
    "- choice of optimizer\n",
    "- hidden layer dimensions -- make sure your choice here makes sense given the number of training examples\n",
    "- batch size\n",
    "- num training steps\n",
    "- (anything else you can think of changing)\n",
    "\n",
    "Do not use the `normalizer_fn` arg on `numeric_column`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "id": "JH6KJp-4-E_Q",
    "outputId": "a1e704a2-addb-48a8-8235-a8baf641c78f",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['symboling', 'normalized-losses', 'wheel-base', 'length', 'width', 'height', 'weight', 'engine-size', 'horsepower', 'peak-rpm', 'city-mpg', 'highway-mpg', 'bore', 'stroke', 'compression-ratio']\n",
      "WARNING:tensorflow:From /Users/eric/.conda/envs/mlbase/lib/python3.8/site-packages/tensorflow/python/util/lazy_loader.py:63: The name tf.estimator.inputs is deprecated. Please use tf.compat.v1.estimator.inputs instead.\n",
      "\n",
      "model_feature_columns [NumericColumn(key='symboling', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='normalized-losses', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='wheel-base', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='length', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='width', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='height', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='weight', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='engine-size', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='horsepower', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='peak-rpm', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='city-mpg', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='highway-mpg', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='bore', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='stroke', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None), NumericColumn(key='compression-ratio', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=None)]\n",
      "INFO:tensorflow:Using default config.\n",
      "WARNING:tensorflow:Using temporary folder as model directory: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_\n",
      "INFO:tensorflow:Using config: {'_model_dir': '/var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true\n",
      "graph_options {\n",
      "  rewrite_options {\n",
      "    meta_optimizer_iterations: ONE\n",
      "  }\n",
      "}\n",
      ", '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}\n",
      "WARNING:tensorflow:From /Users/eric/.conda/envs/mlbase/lib/python3.8/site-packages/tensorflow/python/training/training_util.py:235: Variable.initialized_value (from tensorflow.python.ops.variables) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use Variable.read_value. Variables in 2.X are initialized automatically both in eager and graph (inside tf.defun) contexts.\n",
      "WARNING:tensorflow:From /Users/eric/.conda/envs/mlbase/lib/python3.8/site-packages/tensorflow_estimator/python/estimator/inputs/queues/feeding_queue_runner.py:60: QueueRunner.__init__ (from tensorflow.python.training.queue_runner_impl) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "To construct input pipelines, use the `tf.data` module.\n",
      "WARNING:tensorflow:From /Users/eric/.conda/envs/mlbase/lib/python3.8/site-packages/tensorflow_estimator/python/estimator/inputs/queues/feeding_functions.py:491: add_queue_runner (from tensorflow.python.training.queue_runner_impl) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "To construct input pipelines, use the `tf.data` module.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:From /Users/eric/.conda/envs/mlbase/lib/python3.8/site-packages/tensorflow/python/training/adagrad.py:76: calling Constant.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Call initializer instance with the dtype argument instead of passing it to the constructor\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "WARNING:tensorflow:From /Users/eric/.conda/envs/mlbase/lib/python3.8/site-packages/tensorflow/python/training/monitored_session.py:906: start_queue_runners (from tensorflow.python.training.queue_runner_impl) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "To construct input pipelines, use the `tf.data` module.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...\n",
      "INFO:tensorflow:Saving checkpoints for 0 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...\n",
      "INFO:tensorflow:loss = 2178861600.0, step = 1\n",
      "INFO:tensorflow:global_step/sec: 487.793\n",
      "INFO:tensorflow:loss = 1154110100.0, step = 101 (0.208 sec)\n",
      "INFO:tensorflow:global_step/sec: 711.319\n",
      "INFO:tensorflow:loss = 400658800.0, step = 201 (0.140 sec)\n",
      "INFO:tensorflow:global_step/sec: 768.173\n",
      "INFO:tensorflow:loss = 485674940.0, step = 301 (0.131 sec)\n",
      "INFO:tensorflow:global_step/sec: 703.615\n",
      "INFO:tensorflow:loss = 282755000.0, step = 401 (0.141 sec)\n",
      "INFO:tensorflow:global_step/sec: 591.867\n",
      "INFO:tensorflow:loss = 666994560.0, step = 501 (0.166 sec)\n",
      "INFO:tensorflow:global_step/sec: 612.205\n",
      "INFO:tensorflow:loss = 160398610.0, step = 601 (0.164 sec)\n",
      "INFO:tensorflow:global_step/sec: 806.777\n",
      "INFO:tensorflow:loss = 108026350.0, step = 701 (0.124 sec)\n",
      "INFO:tensorflow:global_step/sec: 744.451\n",
      "INFO:tensorflow:loss = 1367659600.0, step = 801 (0.137 sec)\n",
      "INFO:tensorflow:global_step/sec: 656.043\n",
      "INFO:tensorflow:loss = 553390460.0, step = 901 (0.152 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 1000...\n",
      "INFO:tensorflow:Saving checkpoints for 1000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 1000...\n",
      "INFO:tensorflow:Loss for final step: 714758700.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:36:39Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-1000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.17854s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:36:39\n",
      "INFO:tensorflow:Saving dict for global step 1000: average_loss = 38793990.0, global_step = 1000, label/mean = 12949.43, loss = 611751360.0, prediction/mean = 13033.102\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 1000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-1000\n",
      "scores {'average_loss': 38793990.0, 'label/mean': 12949.43, 'loss': 611751360.0, 'prediction/mean': 13033.102, 'global_step': 1000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-1000\n",
      "WARNING:tensorflow:From /Users/eric/.conda/envs/mlbase/lib/python3.8/site-packages/tensorflow/python/training/saver.py:1077: get_checkpoint_mtimes (from tensorflow.python.training.checkpoint_management) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use standard file utilities to get mtimes.\n",
      "INFO:tensorflow:Running local_init_op.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 1000...\n",
      "INFO:tensorflow:Saving checkpoints for 1000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 1000...\n",
      "INFO:tensorflow:loss = 216372200.0, step = 1001\n",
      "INFO:tensorflow:global_step/sec: 612.771\n",
      "INFO:tensorflow:loss = 257920800.0, step = 1101 (0.165 sec)\n",
      "INFO:tensorflow:global_step/sec: 600.521\n",
      "INFO:tensorflow:loss = 278882620.0, step = 1201 (0.167 sec)\n",
      "INFO:tensorflow:global_step/sec: 585.961\n",
      "INFO:tensorflow:loss = 421352060.0, step = 1301 (0.174 sec)\n",
      "INFO:tensorflow:global_step/sec: 649.834\n",
      "INFO:tensorflow:loss = 542826240.0, step = 1401 (0.151 sec)\n",
      "INFO:tensorflow:global_step/sec: 558.82\n",
      "INFO:tensorflow:loss = 294242560.0, step = 1501 (0.177 sec)\n",
      "INFO:tensorflow:global_step/sec: 601.221\n",
      "INFO:tensorflow:loss = 200007280.0, step = 1601 (0.171 sec)\n",
      "INFO:tensorflow:global_step/sec: 580.37\n",
      "INFO:tensorflow:loss = 470073470.0, step = 1701 (0.169 sec)\n",
      "INFO:tensorflow:global_step/sec: 691.209\n",
      "INFO:tensorflow:loss = 222410530.0, step = 1801 (0.145 sec)\n",
      "INFO:tensorflow:global_step/sec: 663.372\n",
      "INFO:tensorflow:loss = 693115900.0, step = 1901 (0.151 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 2000...\n",
      "INFO:tensorflow:Saving checkpoints for 2000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 2000...\n",
      "INFO:tensorflow:Loss for final step: 157144600.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:36:42Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-2000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.17620s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:36:42\n",
      "INFO:tensorflow:Saving dict for global step 2000: average_loss = 30365810.0, global_step = 2000, label/mean = 12949.43, loss = 478845470.0, prediction/mean = 13050.049\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 2000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-2000\n",
      "scores {'average_loss': 30365810.0, 'label/mean': 12949.43, 'loss': 478845470.0, 'prediction/mean': 13050.049, 'global_step': 2000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-2000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 2000...\n",
      "INFO:tensorflow:Saving checkpoints for 2000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 2000...\n",
      "INFO:tensorflow:loss = 389973630.0, step = 2001\n",
      "INFO:tensorflow:global_step/sec: 517.13\n",
      "INFO:tensorflow:loss = 174584350.0, step = 2101 (0.196 sec)\n",
      "INFO:tensorflow:global_step/sec: 607.681\n",
      "INFO:tensorflow:loss = 457717220.0, step = 2201 (0.165 sec)\n",
      "INFO:tensorflow:global_step/sec: 651.161\n",
      "INFO:tensorflow:loss = 498091420.0, step = 2301 (0.153 sec)\n",
      "INFO:tensorflow:global_step/sec: 620.169\n",
      "INFO:tensorflow:loss = 579908500.0, step = 2401 (0.160 sec)\n",
      "INFO:tensorflow:global_step/sec: 654.798\n",
      "INFO:tensorflow:loss = 251627360.0, step = 2501 (0.155 sec)\n",
      "INFO:tensorflow:global_step/sec: 629.263\n",
      "INFO:tensorflow:loss = 853788200.0, step = 2601 (0.157 sec)\n",
      "INFO:tensorflow:global_step/sec: 672.613\n",
      "INFO:tensorflow:loss = 226052560.0, step = 2701 (0.151 sec)\n",
      "INFO:tensorflow:global_step/sec: 591.142\n",
      "INFO:tensorflow:loss = 309240700.0, step = 2801 (0.171 sec)\n",
      "INFO:tensorflow:global_step/sec: 582.856\n",
      "INFO:tensorflow:loss = 525960200.0, step = 2901 (0.170 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 3000...\n",
      "INFO:tensorflow:Saving checkpoints for 3000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 3000...\n",
      "INFO:tensorflow:Loss for final step: 165793220.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:36:45Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-3000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.19014s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:36:45\n",
      "INFO:tensorflow:Saving dict for global step 3000: average_loss = 26861488.0, global_step = 3000, label/mean = 12949.43, loss = 423585000.0, prediction/mean = 13100.161\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 3000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-3000\n",
      "scores {'average_loss': 26861488.0, 'label/mean': 12949.43, 'loss': 423585000.0, 'prediction/mean': 13100.161, 'global_step': 3000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-3000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 3000...\n",
      "INFO:tensorflow:Saving checkpoints for 3000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 3000...\n",
      "INFO:tensorflow:loss = 741115140.0, step = 3001\n",
      "INFO:tensorflow:global_step/sec: 525.569\n",
      "INFO:tensorflow:loss = 276343700.0, step = 3101 (0.196 sec)\n",
      "INFO:tensorflow:global_step/sec: 615.237\n",
      "INFO:tensorflow:loss = 248371070.0, step = 3201 (0.160 sec)\n",
      "INFO:tensorflow:global_step/sec: 600.229\n",
      "INFO:tensorflow:loss = 383530000.0, step = 3301 (0.166 sec)\n",
      "INFO:tensorflow:global_step/sec: 583.499\n",
      "INFO:tensorflow:loss = 665723650.0, step = 3401 (0.173 sec)\n",
      "INFO:tensorflow:global_step/sec: 578.603\n",
      "INFO:tensorflow:loss = 891271230.0, step = 3501 (0.169 sec)\n",
      "INFO:tensorflow:global_step/sec: 531.318\n",
      "INFO:tensorflow:loss = 813707600.0, step = 3601 (0.191 sec)\n",
      "INFO:tensorflow:global_step/sec: 542.765\n",
      "INFO:tensorflow:loss = 117973780.0, step = 3701 (0.186 sec)\n",
      "INFO:tensorflow:global_step/sec: 554.277\n",
      "INFO:tensorflow:loss = 935103360.0, step = 3801 (0.179 sec)\n",
      "INFO:tensorflow:global_step/sec: 634.421\n",
      "INFO:tensorflow:loss = 1262071300.0, step = 3901 (0.158 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 4000...\n",
      "INFO:tensorflow:Saving checkpoints for 4000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 4000...\n",
      "INFO:tensorflow:Loss for final step: 206941630.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:36:48Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-4000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.19155s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:36:48\n",
      "INFO:tensorflow:Saving dict for global step 4000: average_loss = 25349214.0, global_step = 4000, label/mean = 12949.43, loss = 399737630.0, prediction/mean = 13059.224\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 4000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-4000\n",
      "scores {'average_loss': 25349214.0, 'label/mean': 12949.43, 'loss': 399737630.0, 'prediction/mean': 13059.224, 'global_step': 4000}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-4000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 4000...\n",
      "INFO:tensorflow:Saving checkpoints for 4000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 4000...\n",
      "INFO:tensorflow:loss = 111487816.0, step = 4001\n",
      "INFO:tensorflow:global_step/sec: 648.93\n",
      "INFO:tensorflow:loss = 207377920.0, step = 4101 (0.158 sec)\n",
      "INFO:tensorflow:global_step/sec: 648.844\n",
      "INFO:tensorflow:loss = 215697460.0, step = 4201 (0.156 sec)\n",
      "INFO:tensorflow:global_step/sec: 623.881\n",
      "INFO:tensorflow:loss = 446916700.0, step = 4301 (0.159 sec)\n",
      "INFO:tensorflow:global_step/sec: 642.339\n",
      "INFO:tensorflow:loss = 68027900.0, step = 4401 (0.156 sec)\n",
      "INFO:tensorflow:global_step/sec: 625.118\n",
      "INFO:tensorflow:loss = 223005060.0, step = 4501 (0.161 sec)\n",
      "INFO:tensorflow:global_step/sec: 625.508\n",
      "INFO:tensorflow:loss = 390672320.0, step = 4601 (0.159 sec)\n",
      "INFO:tensorflow:global_step/sec: 732.227\n",
      "INFO:tensorflow:loss = 191996640.0, step = 4701 (0.138 sec)\n",
      "INFO:tensorflow:global_step/sec: 769.603\n",
      "INFO:tensorflow:loss = 754614900.0, step = 4801 (0.129 sec)\n",
      "INFO:tensorflow:global_step/sec: 786.01\n",
      "INFO:tensorflow:loss = 470986180.0, step = 4901 (0.128 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 5000...\n",
      "INFO:tensorflow:Saving checkpoints for 5000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "WARNING:tensorflow:From /Users/eric/.conda/envs/mlbase/lib/python3.8/site-packages/tensorflow/python/training/saver.py:969: remove_checkpoint (from tensorflow.python.training.checkpoint_management) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use standard file APIs to delete files with this prefix.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 5000...\n",
      "INFO:tensorflow:Loss for final step: 177695170.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:36:50Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-5000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.17617s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:36:50\n",
      "INFO:tensorflow:Saving dict for global step 5000: average_loss = 24570766.0, global_step = 5000, label/mean = 12949.43, loss = 387462080.0, prediction/mean = 13138.563\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 5000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-5000\n",
      "scores {'average_loss': 24570766.0, 'label/mean': 12949.43, 'loss': 387462080.0, 'prediction/mean': 13138.563, 'global_step': 5000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-5000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 5000...\n",
      "INFO:tensorflow:Saving checkpoints for 5000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 5000...\n",
      "INFO:tensorflow:loss = 163382160.0, step = 5001\n",
      "INFO:tensorflow:global_step/sec: 644.015\n",
      "INFO:tensorflow:loss = 176191630.0, step = 5101 (0.159 sec)\n",
      "INFO:tensorflow:global_step/sec: 724.028\n",
      "INFO:tensorflow:loss = 91820950.0, step = 5201 (0.138 sec)\n",
      "INFO:tensorflow:global_step/sec: 700.727\n",
      "INFO:tensorflow:loss = 173267940.0, step = 5301 (0.141 sec)\n",
      "INFO:tensorflow:global_step/sec: 725.043\n",
      "INFO:tensorflow:loss = 301676030.0, step = 5401 (0.138 sec)\n",
      "INFO:tensorflow:global_step/sec: 843.716\n",
      "INFO:tensorflow:loss = 357138800.0, step = 5501 (0.119 sec)\n",
      "INFO:tensorflow:global_step/sec: 769.054\n",
      "INFO:tensorflow:loss = 274545120.0, step = 5601 (0.131 sec)\n",
      "INFO:tensorflow:global_step/sec: 747.367\n",
      "INFO:tensorflow:loss = 968657100.0, step = 5701 (0.135 sec)\n",
      "INFO:tensorflow:global_step/sec: 742.083\n",
      "INFO:tensorflow:loss = 76003700.0, step = 5801 (0.134 sec)\n",
      "INFO:tensorflow:global_step/sec: 777.89\n",
      "INFO:tensorflow:loss = 1024511940.0, step = 5901 (0.128 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 6000...\n",
      "INFO:tensorflow:Saving checkpoints for 6000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 6000...\n",
      "INFO:tensorflow:Loss for final step: 150936530.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:36:52Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-6000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.17799s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:36:53\n",
      "INFO:tensorflow:Saving dict for global step 6000: average_loss = 24093442.0, global_step = 6000, label/mean = 12949.43, loss = 379935040.0, prediction/mean = 13030.829\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 6000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-6000\n",
      "scores {'average_loss': 24093442.0, 'label/mean': 12949.43, 'loss': 379935040.0, 'prediction/mean': 13030.829, 'global_step': 6000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-6000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 6000...\n",
      "INFO:tensorflow:Saving checkpoints for 6000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 6000...\n",
      "INFO:tensorflow:loss = 685479900.0, step = 6001\n",
      "INFO:tensorflow:global_step/sec: 633.4\n",
      "INFO:tensorflow:loss = 441601630.0, step = 6101 (0.162 sec)\n",
      "INFO:tensorflow:global_step/sec: 779.321\n",
      "INFO:tensorflow:loss = 416646530.0, step = 6201 (0.128 sec)\n",
      "INFO:tensorflow:global_step/sec: 763.399\n",
      "INFO:tensorflow:loss = 121653380.0, step = 6301 (0.132 sec)\n",
      "INFO:tensorflow:global_step/sec: 740.725\n",
      "INFO:tensorflow:loss = 158561460.0, step = 6401 (0.134 sec)\n",
      "INFO:tensorflow:global_step/sec: 747.664\n",
      "INFO:tensorflow:loss = 1206589400.0, step = 6501 (0.134 sec)\n",
      "INFO:tensorflow:global_step/sec: 729.112\n",
      "INFO:tensorflow:loss = 544550660.0, step = 6601 (0.137 sec)\n",
      "INFO:tensorflow:global_step/sec: 727.643\n",
      "INFO:tensorflow:loss = 113839950.0, step = 6701 (0.136 sec)\n",
      "INFO:tensorflow:global_step/sec: 802.188\n",
      "INFO:tensorflow:loss = 458637000.0, step = 6801 (0.124 sec)\n",
      "INFO:tensorflow:global_step/sec: 705.249\n",
      "INFO:tensorflow:loss = 405443360.0, step = 6901 (0.141 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 7000...\n",
      "INFO:tensorflow:Saving checkpoints for 7000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 7000...\n",
      "INFO:tensorflow:Loss for final step: 1130919400.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:36:55Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-7000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.17586s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:36:55\n",
      "INFO:tensorflow:Saving dict for global step 7000: average_loss = 23814748.0, global_step = 7000, label/mean = 12949.43, loss = 375540260.0, prediction/mean = 12887.202\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 7000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-7000\n",
      "scores {'average_loss': 23814748.0, 'label/mean': 12949.43, 'loss': 375540260.0, 'prediction/mean': 12887.202, 'global_step': 7000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-7000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 7000...\n",
      "INFO:tensorflow:Saving checkpoints for 7000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 7000...\n",
      "INFO:tensorflow:loss = 385394560.0, step = 7001\n",
      "INFO:tensorflow:global_step/sec: 632.88\n",
      "INFO:tensorflow:loss = 444891400.0, step = 7101 (0.162 sec)\n",
      "INFO:tensorflow:global_step/sec: 720.42\n",
      "INFO:tensorflow:loss = 140703740.0, step = 7201 (0.137 sec)\n",
      "INFO:tensorflow:global_step/sec: 758.259\n",
      "INFO:tensorflow:loss = 495819330.0, step = 7301 (0.133 sec)\n",
      "INFO:tensorflow:global_step/sec: 752.224\n",
      "INFO:tensorflow:loss = 806502100.0, step = 7401 (0.132 sec)\n",
      "INFO:tensorflow:global_step/sec: 787.29\n",
      "INFO:tensorflow:loss = 419060770.0, step = 7501 (0.128 sec)\n",
      "INFO:tensorflow:global_step/sec: 678.794\n",
      "INFO:tensorflow:loss = 612093400.0, step = 7601 (0.145 sec)\n",
      "INFO:tensorflow:global_step/sec: 702.641\n",
      "INFO:tensorflow:loss = 181521220.0, step = 7701 (0.143 sec)\n",
      "INFO:tensorflow:global_step/sec: 760.458\n",
      "INFO:tensorflow:loss = 132836010.0, step = 7801 (0.134 sec)\n",
      "INFO:tensorflow:global_step/sec: 753.948\n",
      "INFO:tensorflow:loss = 87113130.0, step = 7901 (0.132 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 8000...\n",
      "INFO:tensorflow:Saving checkpoints for 8000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 8000...\n",
      "INFO:tensorflow:Loss for final step: 227725600.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:36:57Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-8000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.17561s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:36:57\n",
      "INFO:tensorflow:Saving dict for global step 8000: average_loss = 23521448.0, global_step = 8000, label/mean = 12949.43, loss = 370915140.0, prediction/mean = 13145.544\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 8000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-8000\n",
      "scores {'average_loss': 23521448.0, 'label/mean': 12949.43, 'loss': 370915140.0, 'prediction/mean': 13145.544, 'global_step': 8000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-8000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 8000...\n",
      "INFO:tensorflow:Saving checkpoints for 8000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 8000...\n",
      "INFO:tensorflow:loss = 124635570.0, step = 8001\n",
      "INFO:tensorflow:global_step/sec: 629.39\n",
      "INFO:tensorflow:loss = 127904050.0, step = 8101 (0.161 sec)\n",
      "INFO:tensorflow:global_step/sec: 751.716\n",
      "INFO:tensorflow:loss = 769458600.0, step = 8201 (0.134 sec)\n",
      "INFO:tensorflow:global_step/sec: 752.491\n",
      "INFO:tensorflow:loss = 262612210.0, step = 8301 (0.133 sec)\n",
      "INFO:tensorflow:global_step/sec: 758.875\n",
      "INFO:tensorflow:loss = 201513600.0, step = 8401 (0.132 sec)\n",
      "INFO:tensorflow:global_step/sec: 742.24\n",
      "INFO:tensorflow:loss = 193429900.0, step = 8501 (0.136 sec)\n",
      "INFO:tensorflow:global_step/sec: 651.887\n",
      "INFO:tensorflow:loss = 1159808400.0, step = 8601 (0.151 sec)\n",
      "INFO:tensorflow:global_step/sec: 701.951\n",
      "INFO:tensorflow:loss = 927779600.0, step = 8701 (0.144 sec)\n",
      "INFO:tensorflow:global_step/sec: 673.387\n",
      "INFO:tensorflow:loss = 134751500.0, step = 8801 (0.147 sec)\n",
      "INFO:tensorflow:global_step/sec: 723.234\n",
      "INFO:tensorflow:loss = 451980100.0, step = 8901 (0.139 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 9000...\n",
      "INFO:tensorflow:Saving checkpoints for 9000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 9000...\n",
      "INFO:tensorflow:Loss for final step: 114730430.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:37:00Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-9000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.19055s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:37:00\n",
      "INFO:tensorflow:Saving dict for global step 9000: average_loss = 23303732.0, global_step = 9000, label/mean = 12949.43, loss = 367481920.0, prediction/mean = 13042.049\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 9000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-9000\n",
      "scores {'average_loss': 23303732.0, 'label/mean': 12949.43, 'loss': 367481920.0, 'prediction/mean': 13042.049, 'global_step': 9000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-9000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 9000...\n",
      "INFO:tensorflow:Saving checkpoints for 9000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 9000...\n",
      "INFO:tensorflow:loss = 649782000.0, step = 9001\n",
      "INFO:tensorflow:global_step/sec: 558.248\n",
      "INFO:tensorflow:loss = 753044860.0, step = 9101 (0.182 sec)\n",
      "INFO:tensorflow:global_step/sec: 674.309\n",
      "INFO:tensorflow:loss = 289949820.0, step = 9201 (0.149 sec)\n",
      "INFO:tensorflow:global_step/sec: 608.35\n",
      "INFO:tensorflow:loss = 147997860.0, step = 9301 (0.166 sec)\n",
      "INFO:tensorflow:global_step/sec: 659.866\n",
      "INFO:tensorflow:loss = 397985280.0, step = 9401 (0.150 sec)\n",
      "INFO:tensorflow:global_step/sec: 596.332\n",
      "INFO:tensorflow:loss = 120440030.0, step = 9501 (0.168 sec)\n",
      "INFO:tensorflow:global_step/sec: 620.332\n",
      "INFO:tensorflow:loss = 70000750.0, step = 9601 (0.164 sec)\n",
      "INFO:tensorflow:global_step/sec: 613.72\n",
      "INFO:tensorflow:loss = 158345440.0, step = 9701 (0.161 sec)\n",
      "INFO:tensorflow:global_step/sec: 616.389\n",
      "INFO:tensorflow:loss = 542189200.0, step = 9801 (0.161 sec)\n",
      "INFO:tensorflow:global_step/sec: 581.156\n",
      "INFO:tensorflow:loss = 898998140.0, step = 9901 (0.173 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 10000...\n",
      "INFO:tensorflow:Saving checkpoints for 10000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 10000...\n",
      "INFO:tensorflow:Loss for final step: 189728640.0.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:37:02Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-10000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.19125s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:37:03\n",
      "INFO:tensorflow:Saving dict for global step 10000: average_loss = 23052622.0, global_step = 10000, label/mean = 12949.43, loss = 363522100.0, prediction/mean = 13146.191\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 10000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-10000\n",
      "scores {'average_loss': 23052622.0, 'label/mean': 12949.43, 'loss': 363522100.0, 'prediction/mean': 13146.191, 'global_step': 10000}\n"
     ]
    }
   ],
   "source": [
    "# This code \"works\", but because of bad hyperparameter choices it gets NaN loss\n",
    "# during training. Try fixing this.\n",
    "\n",
    "batch_size = 16\n",
    "\n",
    "print(numeric_feature_names)\n",
    "x_df = car_data[numeric_feature_names]\n",
    "y_series = car_data['price']\n",
    "\n",
    "# Create input_fn's so that the estimator knows how to read in your data.\n",
    "train_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    batch_size=batch_size,\n",
    "    num_epochs=None,\n",
    "    shuffle=True)\n",
    "\n",
    "eval_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    batch_size=batch_size,\n",
    "    shuffle=False)\n",
    "\n",
    "predict_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    batch_size=batch_size,\n",
    "    shuffle=False)\n",
    "\n",
    "# Feature columns allow the model to parse the data, perform common\n",
    "# preprocessing, and automatically generate an input layer for the tf.Estimator.\n",
    "model_feature_columns = [\n",
    "    tf.feature_column.numeric_column(feature_name) for feature_name in numeric_feature_names\n",
    "]\n",
    "print('model_feature_columns', model_feature_columns)\n",
    "\n",
    "est = tf.estimator.DNNRegressor(\n",
    "    feature_columns=model_feature_columns,\n",
    "    hidden_units=[64],\n",
    "    optimizer=tf.train.AdagradOptimizer(learning_rate=0.01),\n",
    "  )\n",
    "\n",
    "# TRAIN\n",
    "num_print_statements = 10\n",
    "num_training_steps = 10000\n",
    "for _ in range(num_print_statements):\n",
    "  est.train(train_input_fn, steps=num_training_steps // num_print_statements)\n",
    "  scores = est.evaluate(eval_input_fn)\n",
    "  \n",
    "  # The `scores` dictionary has several metrics automatically generated by the\n",
    "  # canned Estimator.\n",
    "  # `average_loss` is the average loss for an individual example.\n",
    "  # `loss` is the summed loss for the batch.\n",
    "  # In addition to these scalar losses, you may find the visualization functions\n",
    "  # in the next cell helpful for debugging model quality.\n",
    "  print('scores', scores)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Rxto3DwsjYw4"
   },
   "source": [
    "### Visualize your model's predictions\n",
    "\n",
    "After you have a trained model, it may be helpful to understand how your model's inference differs from the actual data.\n",
    "\n",
    "This helper function `scatter_plot_inference` does that for you. Real data is in grey. Your model's predictions are in orange.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "uGSXwX2fju1N",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpdgn4ymp_/model.ckpt-10000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 972x1620 with 15 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "\n",
    "\n",
    "def scatter_plot_inference_grid(est, x_df, feature_names):\n",
    "  \"\"\"Plots the predictions of the model against each feature.\n",
    "  \n",
    "  Args:\n",
    "    est: The trained tf.Estimator.\n",
    "    x_df: The pandas dataframe with the input data (used to create\n",
    "      predict_input_fn).\n",
    "    feature_names: An iterable of string feature names to plot.\n",
    "  \"\"\"\n",
    "  def scatter_plot_inference(axis,\n",
    "                             x_axis_feature_name,\n",
    "                             y_axis_feature_name,\n",
    "                             predictions):\n",
    "    \"\"\"Generate one subplot.\"\"\"\n",
    "    # Plot the real data in grey.\n",
    "    y_axis_feature_name = 'price'\n",
    "    axis.set_ylabel(y_axis_feature_name)\n",
    "    axis.set_xlabel(x_axis_feature_name)\n",
    "    axis.scatter(car_data[x_axis_feature_name],\n",
    "                 car_data[y_axis_feature_name],\n",
    "                 c='grey')\n",
    "\n",
    "    # Plot the predicted data in orange.\n",
    "    axis.scatter(car_data[x_axis_feature_name], predictions, c='orange')\n",
    "\n",
    "  predict_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    batch_size=batch_size,\n",
    "    shuffle=False)\n",
    "\n",
    "  predictions = [\n",
    "    x['predictions'][0]\n",
    "    for x in est.predict(predict_input_fn)\n",
    "  ]\n",
    "  \n",
    "  num_cols = 3\n",
    "  num_rows = int(math.ceil(len(feature_names)/float(num_cols)))\n",
    "  f, axarr = plt.subplots(num_rows, num_cols)\n",
    "  size = 4.5\n",
    "  f.set_size_inches(num_cols*size, num_rows*size)\n",
    "  \n",
    "  for i, feature_name in enumerate(numeric_feature_names):\n",
    "    axis = axarr[int(i/num_cols), i%num_cols]\n",
    "    scatter_plot_inference(axis, feature_name, 'price', predictions)\n",
    "  plt.show()\n",
    "\n",
    "scatter_plot_inference_grid(est, x_df, numeric_feature_names)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "uBZI8f_8Yfph"
   },
   "source": [
    "## Task 2: Take your best numeric model from earlier. Add normalization.\n",
    "\n",
    "### Add normalization to your best numeric model from earlier\n",
    "\n",
    "- You decide what type of normalization to add, and for which features\n",
    "- You will need to use the `normalizer_fn` arg on [`numeric_column`](https://g3doc.corp.google.com/learning/brain/public/g3doc/api_docs/python/tf/feature_column/numeric_column.md?cl=head)\n",
    "    - An example of a silly normalizer_fn that shifts inputs down by 1, and then negates the value:\n",
    "    \n",
    "         normalizer_fn = lambda x: tf.neg(tf.subtract(x, 1))\n",
    "\n",
    "- You may find these pandas functions helpful:\n",
    "    - dataframe.mean()['your_feature_name']\n",
    "    - dataframe.std()['your_feature_name']\n",
    "- You will need to retune the hyperparameters from earlier.\n",
    "\n",
    "\n",
    "**Does normalization improve model quality on this dataset? Why or why not?**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "jY_C_QgcZg1-",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXMAAAEICAYAAACtXxSQAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/d3fzzAAAACXBIWXMAAAsTAAALEwEAmpwYAAASl0lEQVR4nO3dfZBddX3H8fcXEE1ZhaTANgbq0hYZGTOg2aozaNkFtQjWYCsODjphSietIxadUBuKY7Uda2qLnXbGVtPqmFHrmiqUKDNVTFkZZ3xKKBiYgPiQhgTcjJhQ1jLYyLd/7Ilcw+69Z3fv3Xvuz/drZueeh98955M7m8+ee/bcs5GZSJIG2zH9DiBJWjzLXJIKYJlLUgEsc0kqgGUuSQWwzCWpAJa5BlpE7ImIly/xPkciIiPiuKXcr9SOZS510I8fGNJ8WeaSVADLXEWIiGMiYmNEfDciHo6IrRGxolp35LTIuojYGxE/jIjrW567LCK2RMTBiNgdEe+IiH3Vuo8Dvwp8LiKmI+IdLbu9YrbtSf1gmasUfwxcCpwPPBs4CHzwqDEvBc4CLgTeFRHPq5b/OTAC/BrwCuCNR56QmW8C9gK/k5lDmfn+GtuTlpxlrlL8IXB9Zu7LzMeBdwOvO+qXlO/JzMcy8y7gLuCcavnrgb/KzIOZuQ/4h5r7nGt70pLzt/EqxXOAmyLiiZZlPwWGW+Z/0DL9v8BQNf1s4IGWda3T7cy1PWnJeWSuUjwAvCozT2r5ekZm7q/x3IeA01rmTz9qvbcWVeNZ5irFh4D3RsRzACLilIhYW/O5W4HrImJ5RKwCrj5q/RQz59OlxrLMVYq/B7YBX4yIR4GvAS+u+dy/APYB3we+BHwGeLxl/fuAd0bEoYi4tnuRpe4J/ziF9PMi4s3A5Zl5fr+zSHV5ZK5feBGxMiLOq65VPwvYANzU71zSfHg1iwTHAx8GzgAOARPAP/YzkDRfnmaRpAJ4mkWSCrCkp1lOPvnkHBkZ6Tjuxz/+MSeccELvA3XZIOY289IZxNyDmBkGM3e7zDt37vxhZp7SdgOZuWRfa9asyTpuu+22WuOaZhBzm3npDGLuQcycOZi522UGdmSHfvU0iyQVwDKXpAJY5pJUAMtckgpgmUtSASxzSSqAZS5JBbDMJakAlrkkFcC7JqpRRjbe0nHMhtWHubLGuPnYs+mSrm5PWmoemUtSASxzSSqAZS5JBbDMJakAlrkkFcAyl6QCWOaSVADLXJIKYJlLUgEsc0kqgGUuSQWodW+WiNgDPAr8FDicmaMRsQL4NDAC7AFen5kHexNTktTOfI7MxzPz3MwcreY3Atsz80xgezUvSeqDxZxmWQtsqaa3AJcuOo0kaUEiMzsPivg+cBBI4MOZuTkiDmXmSS1jDmbm8lmeux5YDzA8PLxmYmKi4/6mp6cZGhqq/Y9oikHM3bTMu/Y/0nHM8DKYeqy7+1296sTubnAWTXut6xjEzDCYudtlHh8f39lyVmRWde9nfl5mPhgRpwK3RsS9dQNm5mZgM8Do6GiOjY11fM7k5CR1xjXNIOZuWuY69ynfsPowN+zq7q3491wx1tXtzaZpr3Udg5gZBjP3YjPXOs2SmQ9WjweAm4AXAVMRsRKgejyw4BSSpEXpWOYRcUJEPPPINPBK4G5gG7CuGrYOuLlXISVJ7dV5rzoM3BQRR8b/a2b+R0R8E9gaEVcBe4HLehdTktROxzLPzO8B58yy/GHgwl6EkiTNj58AlaQCWOaSVADLXJIKYJlLUgEsc0kqgGUuSQWwzCWpAJa5JBXAMpekAljmklQAy1ySCmCZS1IBLHNJKoBlLkkFsMwlqQCWuSQVwDKXpAJY5pJUAMtckgpgmUtSASxzSSqAZS5JBbDMJakAlrkkFcAyl6QCWOaSVADLXJIKYJlLUgEsc0kqQO0yj4hjI+K/IuLz1fyKiLg1Iu6vHpf3LqYkqZ35HJlfA+xumd8IbM/MM4Ht1bwkqQ9qlXlEnAZcAvxLy+K1wJZqegtwaVeTSZJqi8zsPCjiM8D7gGcC12bmqyPiUGae1DLmYGY+5VRLRKwH1gMMDw+vmZiY6Li/6elphoaGav8jmmIQczct8679j3QcM7wMph7r7n5XrzqxuxucRdNe6zoGMTMMZu52mcfHx3dm5mi75x/XaQcR8WrgQGbujIix+QbMzM3AZoDR0dEcG+u8icnJSeqMa5pBzN20zFduvKXjmA2rD3PDro7fuvOy54qxrm5vNk17resYxMwwmLkXm7nO/4jzgNdExMXAM4BnRcQngKmIWJmZD0XESuDAglNIkhal4znzzLwuM0/LzBHgcuA/M/ONwDZgXTVsHXBzz1JKktpazHvVTcDWiLgK2Atc1p1I0tIbqXF6Z7E2rD4862mkPZsu6fm+Vb55lXlmTgKT1fTDwIXdjyRJmi8/ASpJBbDMJakAlrkkFcAyl6QCWOaSVADLXJIKYJlLUgEsc0kqgGUuSQWwzCWpAJa5JBXAMpekAljmklQAy1ySCmCZS1IBLHNJKoBlLkkFsMwlqQCWuSQVwDKXpAJY5pJUAMtckgpgmUtSASxzSSqAZS5JBbDMJakAlrkkFcAyl6QCWOaSVADLXJIK0LHMI+IZEfGNiLgrIu6JiPdUy1dExK0RcX/1uLz3cSVJs6lzZP44cEFmngOcC1wUES8BNgLbM/NMYHs1L0nqg45lnjOmq9mnVV8JrAW2VMu3AJf2IqAkqbPIzM6DIo4FdgK/AXwwM/80Ig5l5kktYw5m5lNOtUTEemA9wPDw8JqJiYmO+5uenmZoaKj2P6IpBjF30zLv2v9IxzHDy2DqsSUI02Vz5V696sSlD1NT074/6hrE3O0yj4+P78zM0XbPr1XmPxsccRJwE/BW4Ct1yrzV6Oho7tixo+N+JicnGRsbq52rKQYxd9Myj2y8peOYDasPc8Ou45YgTXfNlXvPpkv6kKaepn1/1DWIudtljoiOZT6vq1ky8xAwCVwETEXEympHK4ED89mWJKl76lzNckp1RE5ELANeDtwLbAPWVcPWATf3KKMkqYM671VXAluq8+bHAFsz8/MR8VVga0RcBewFLuthzl9IdU45LNaG1Ye58qj9NPltv6TZdSzzzPwW8IJZlj8MXNiLUJKk+fEToJJUAMtckgpgmUtSASxzSSqAZS5JBbDMJakAlrkkFcAyl6QCWOaSVADLXJIKYJlLUgEsc0kqgGUuSQWwzCWpAJa5JBXAMpekAljmklQAy1ySCmCZS1IBLHNJKoBlLkkFOK7fAdQ8Ixtv6XeEXyj9er33bLqkL/tVb3hkLkkFsMwlqQCWuSQVwDKXpAJY5pJUAMtckgpgmUtSATqWeUScHhG3RcTuiLgnIq6plq+IiFsj4v7qcXnv40qSZlPnyPwwsCEznwe8BHhLRJwNbAS2Z+aZwPZqXpLUBx3LPDMfysw7qulHgd3AKmAtsKUatgW4tEcZJUkdRGbWHxwxAtwOPB/Ym5kntaw7mJlPOdUSEeuB9QDDw8NrJiYmOu5nenqaoaGh2rmaotu5d+1/pGvbmsvwMph6rOe76apBzAyDmbtXmVevOrH7G20xiB3SLvP4+PjOzBxt9/zaZR4RQ8CXgfdm5o0RcahOmbcaHR3NHTt2dNzX5OQkY2NjtXI1SbdzL8U9OzasPswNuwbrFj2DmBkGM3evMvf6vjCD2CHtMkdExzKvdTVLRDwN+Czwycy8sVo8FRErq/UrgQN1Q0uSuqvO1SwBfATYnZkfaFm1DVhXTa8Dbu5+PElSHXXeP50HvAnYFRF3Vsv+DNgEbI2Iq4C9wGU9SShJ6qhjmWfmV4CYY/WF3Y0jSVoIPwEqSQWwzCWpAJa5JBXAMpekAljmklQAy1ySCmCZS1IBLHNJKoBlLkkFsMwlqQCWuSQVwDKXpAJY5pJUAMtckgpgmUtSASxzSSqAZS5JBbDMJakAlrkkFcAyl6QCWOaSVADLXJIKYJlLUgEsc0kqgGUuSQWwzCWpAJa5JBXAMpekAljmklSAjmUeER+NiAMRcXfLshURcWtE3F89Lu9tTElSO3WOzD8GXHTUso3A9sw8E9hezUuS+qRjmWfm7cCPjlq8FthSTW8BLu1uLEnSfERmdh4UMQJ8PjOfX80fysyTWtYfzMxZT7VExHpgPcDw8PCaiYmJjvubnp5maGioTv5G6XbuXfsf6dq25jK8DKYe6/luumoQM8Ng5u5V5tWrTuz+RlsMYoe0yzw+Pr4zM0fbPf+4nqRqkZmbgc0Ao6OjOTY21vE5k5OT1BnXNN3OfeXGW7q2rblsWH2YG3b1/NugqwYxMwxm7l5l3nPFWNe32WoQO2SxmRd6NctURKwEqB4PLDiBJGnRFlrm24B11fQ64ObuxJEkLUSdSxM/BXwVOCsi9kXEVcAm4BURcT/wimpektQnHU+GZeYb5lh1YZezSJIWyE+ASlIBLHNJKsBgXSclqQgjPb7sdsPqw7Ne2rtn0yU93W8/eWQuSQWwzCWpAJa5JBXAMpekAljmklQAy1ySCmCZS1IBLHNJKoBlLkkFsMwlqQCWuSQVwDKXpAJY5pJUAMtckgpgmUtSASxzSSqAZS5JBbDMJakAA/Nn43r9Z6a6Ya4/VSWpGfrZI73+k3UemUtSASxzSSqAZS5JBbDMJakAlrkkFcAyl6QCWOaSVADLXJIKsKgyj4iLIuK+iPhORGzsVihJ0vwsuMwj4ljgg8CrgLOBN0TE2d0KJkmqbzFH5i8CvpOZ38vMnwATwNruxJIkzUdk5sKeGPE64KLM/INq/k3AizPz6qPGrQfWV7NnAffV2PzJwA8XFKy/BjG3mZfOIOYexMwwmLnbZX5OZp7S7smLudFWzLLsKT8ZMnMzsHleG47YkZmjCw3WL4OY28xLZxBzD2JmGMzci828mNMs+4DTW+ZPAx5cxPYkSQu0mDL/JnBmRJwREccDlwPbuhNLkjQfCz7NkpmHI+Jq4AvAscBHM/OeLuWa12mZBhnE3GZeOoOYexAzw2DmXlTmBf8CVJLUHH4CVJIKYJlLUgH6UuYR8dGIOBARd8+y7tqIyIg4uWXZddUtA+6LiN9e2rQ/y/CUzBHx7ojYHxF3Vl8XNylzlWPW1zoi3lpluyci3t+yvO+553itP93yOu+JiDsHIPO5EfG1KvOOiHhRkzJXOWbLfU5EfDUidkXE5yLiWS3r+p47Ik6PiNsiYnf1/XtNtXxFRNwaEfdXj8ubkrtN5suq+SciYvSo58wvc2Yu+RfwW8ALgbuPWn46M79Q/W/g5GrZ2cBdwNOBM4DvAsc2ITPwbuDaWcY2InOb3OPAl4CnV/OnNin3XN8fLetvAN7V9MzAF4FXVdMXA5NNytwm9zeB86vp3wf+skm5gZXAC6vpZwLfrrK9H9hYLd8I/HVTcrfJ/DxmPkw5CYy2jJ935r4cmWfm7cCPZln1d8A7+PkPH60FJjLz8cz8PvAdZm4lsKTaZJ5NIzLDnLnfDGzKzMerMQeq5Y3I3e61jogAXg98qlrU5MwJHDmqPZEnP4fRiMwwZ+6zgNur6VuB36umG5E7Mx/KzDuq6UeB3cCqKt+WatgW4NJquu+558qcmbszc7ZPxc87c2POmUfEa4D9mXnXUatWAQ+0zO+rljXF1RHxrert6pG3dU3P/FzgZRHx9Yj4ckT8ZrW86bkBXgZMZeb91XyTM78N+JuIeAD4W+C6anmTMwPcDbymmr6MJz8c2LjcETECvAD4OjCcmQ/BTHkCp1bDGpX7qMxzmXfmRpR5RPwScD3wrtlWz7KsKddT/hPw68C5wEPMvP2HZmeGmc8XLAdeAvwJsLU64m16boA38ORROTQ785uBt2fm6cDbgY9Uy5ucGWZOrbwlInYyc0rgJ9XyRuWOiCHgs8DbMvN/2g2dZVlfcvcycyPKnJlCPAO4KyL2MHNrgDsi4ldo8G0DMnMqM3+amU8A/8yTb4Mam7myD7gxZ3wDeIKZm/w0OndEHAf8LvDplsVNzrwOuLGa/jcG5PsjM+/NzFdm5hpmfnB+t1rVmNwR8TRmSvGTmXnkNZ6KiJXV+pXAkdOHjcg9R+a5zDtzI8o8M3dl5qmZOZKZI8z8Q16YmT9g5hYBl0fE0yPiDOBM4Bt9jPszR75xKq9l5u0pNDhz5d+BCwAi4rnA8czcra3puV8O3JuZ+1qWNTnzg8D51fQFwJFTQ03OTEScWj0eA7wT+FC1qhG5q3eRHwF2Z+YHWlZtY+YHKNXjzS3L+5q7Tea5zD/zUv8muvpN7aeYOS3xf8wU91VHrd9DdTVLNX89M0cH91FdHdCEzMDHgV3At6oXf2WTMrfJfTzwCWZ++NwBXNCk3HN9fwAfA/5olvGNzAy8FNjJzFUJXwfWNClzm9zXMHO1xbeBTVSfFG9K7up1zer/3Z3V18XALwPbmfmhuR1Y0ZTcbTK/tnrdHwemgC8sNLMf55ekAjTiNIskaXEsc0kqgGUuSQWwzCWpAJa5JBXAMpekAljmklSA/wd/6MePEFIymAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAEICAYAAACktLTqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/d3fzzAAAACXBIWXMAAAsTAAALEwEAmpwYAAASyUlEQVR4nO3df7DldX3f8edLEEWusrvB3Nks1CXJ1paw1sCNQGzsXYnRiClkps7gqFkSzLZTktCWNrMknTp2yoS0Q2o6pB0xaDajsiGEFkbGRrLmxjE/MLuKWWClC7qBBbKrlSWuYTSL7/5xvkuOy71w7/neX+fT52Pmzvmez/fX68LZ1/3ezzn3nFQVkqS2vGilA0iSFp/lLkkNstwlqUGWuyQ1yHKXpAZZ7pLUIMtdApJckeQzK51DWiyWuyQ1yHKXekpy8kpnkE5kuWvsJDmQ5NokDyR5MsmHk7y0W/e2JPcmOZLkT5K8Zmi/7UkeTvL1bt+ffJ5z/Jckn0ly+hzrK8lVSfYD+4fGfiHJl5J8tTvGi7p1VyT54yT/tcv2pSQ/3I0/muRwkq2L+h9K/1+z3DWu3gm8Gfg+4O8D/z7JecCHgH8OfBfwAeDOJC/p9nkY+BHgdOB9wEeSrB8+aJIXJfkg8Brgx6rqqefJcBlwAXDO0NhPAlPAecClwM8MrbsA+Isu28eAncAPAd8PvAu4McnE/P8TSHOz3DWubqyqR6vqa8B1wDuAnwU+UFX3VNUzVbUD+CZwIUBV/W5VPV5V366q32Fwxf26oWO+GLgFWAf8RFX9zQtk+JWq+lpVPT009qvd2CPA+7tcx325qj5cVc8AvwOcBfzHqvpmVX0S+BaDopd6s9w1rh4dWv5L4HuAVwHXdNMeR5IcYVCg3wOQ5KeGpmyOAOcCZwwd5/sZXG2/r6q+dXwwyf1JjnZfPzJHhufLddyhoeWnAarqxDGv3LUofCJI4+qsoeW/BzzOoFivq6rrTtw4yauADwIXA39aVc8kuRfI0Gb7gN8APpHkjVX1IEBV/cAcGWZ7S9WzgPtPyCUtO6/cNa6uSnJmknXALzGY5vgg8C+SXJCB05JckuTlwGkMyvgrAEl+msGV+3eoqlu64/1Bku8bIde/S7I2yVnA1V0uadl55a5x9THgkwymPe4A/lNV/U2SnwVuBDYxmOb4DPDpqnogyQ3AnwLfBn4b+OPZDlxVO5KcAnwqyT+pqgMLyHUHsIfBk7a/Bdw8wvcm9RY/rEPjJskB4D1V9QcrnWVYkgI2VdVDK51FclpGkhpkuUtSg5yWkaQGeeUuSQ1aFa+WOeOMM2rjxo0j7/+Nb3yD0047bfECLRNzLy9zLy9zL709e/Z8tapeOdu6VVHuGzduZPfu3SPvPzMzw/T09OIFWibmXl7mXl7mXnpJ/nKudU7LSFKDLHdJapDlLkkNstwlqUGWuyQ1yHKXpAZZ7pLUIMtdkhpkuUtSg1bFX6hKWl02br+r9zGu2XyMKxZ4nAPXX9L7vBp4wSv3JB9KcjjJfUNj65LcnWR/d7t2aN21SR5K8mCSNy9VcEnS3OYzLfNbwFtOGNsO7KqqTcCu7j5JzgEuB36g2+e/Jzlp0dJKkublBcu9qj4NfO2E4UuBHd3yDuCyofGdVfXNqvoy8BDwusWJKkmar3l9WEeSjcDHq+rc7v6RqloztP7Jqlqb5Ebgz6rqI934zcAnquq2WY65DdgGMDk5ef7OnTtH/iaOHj3KxMTEyPuvFHMvL3PP397Hnup9jMlT4dDTC9tn84bTe5+3r3F6nGzZsmVPVU3Ntm6xn1DNLGOz/vSoqpuAmwCmpqaqz1tsjtNbdA4z9/Iy9/wt9InQ2Vyz+Rg37F1YxRx453Tv8/Y1ro+TE436UshDSdYDdLeHu/GDwFlD250JPD56PEnSKEYt9zuBrd3yVuCOofHLk7wkydnAJuCz/SJKkhbqBX9nSnILMA2ckeQg8F7geuDWJFcCjwBvB6iq+5PcCjwAHAOuqqpnlii7JGkOL1juVfWOOVZdPMf21wHX9QklSerHtx+QpAZZ7pLUIMtdkhpkuUtSgyx3SWqQ5S5JDbLcJalBlrskNchyl6QGWe6S1CDLXZIaZLlLUoMsd0lqkOUuSQ2y3CWpQZa7JDXIcpekBlnuktQgy12SGmS5S1KDLHdJapDlLkkNstwlqUGWuyQ1yHKXpAZZ7pLUIMtdkhpkuUtSgyx3SWqQ5S5JDbLcJalBvco9yb9Ocn+S+5LckuSlSdYluTvJ/u527WKFlSTNz8jlnmQD8AvAVFWdC5wEXA5sB3ZV1SZgV3dfkrSM+k7LnAycmuRk4GXA48ClwI5u/Q7gsp7nkCQtUKpq9J2Tq4HrgKeBT1bVO5Mcqao1Q9s8WVXPmZpJsg3YBjA5OXn+zp07R85x9OhRJiYmRt5/pZh7eZl7/vY+9lTvY0yeCoeeXtg+mzec3vu8fY3T42TLli17qmpqtnUnj3rQbi79UuBs4Ajwu0neNd/9q+om4CaAqampmp6eHjUKMzMz9Nl/pZh7eZl7/q7YflfvY1yz+Rg37F1YxRx453Tv8/Y1ro+TE/WZlvlR4MtV9ZWq+lvgduCHgUNJ1gN0t4f7x5QkLUSfcn8EuDDJy5IEuBjYB9wJbO222Qrc0S+iJGmhRp6Wqap7ktwGfA44BnyewTTLBHBrkisZ/AB4+2IElSTN38jlDlBV7wXee8LwNxlcxUuSVoh/oSpJDbLcJalBlrskNchyl6QGWe6S1CDLXZIaZLlLUoMsd0lqkOUuSQ2y3CWpQZa7JDXIcpekBlnuktQgy12SGmS5S1KDLHdJapDlLkkNstwlqUGWuyQ1yHKXpAZZ7pLUIMtdkhpkuUtSgyx3SWqQ5S5JDbLcJalBlrskNchyl6QGWe6S1KCTVzqAJB23cftdK3LeA9dfsiLnXUpeuUtSgyx3SWpQr3JPsibJbUm+mGRfkouSrEtyd5L93e3axQorSZqfvlfuvw7876r6B8A/AvYB24FdVbUJ2NXdlyQto5HLPckrgDcANwNU1beq6ghwKbCj22wHcFm/iJKkhUpVjbZj8lrgJuABBlfte4Crgceqas3Qdk9W1XOmZpJsA7YBTE5Onr9z586RcgAcPXqUiYmJkfdfKeZeXuaev72PPdX7GJOnwqGnFyHMMti84fRnl8fpcbJly5Y9VTU127o+5T4F/Bnw+qq6J8mvA38N/Px8yn3Y1NRU7d69e6QcADMzM0xPT4+8/0ox9/Iy9/wtxksSr9l8jBv2jserrYdfCjlOj5Mkc5Z7nzn3g8DBqrqnu38bcB5wKMn67sTrgcM9ziFJGsHI5V5VfwU8muTV3dDFDKZo7gS2dmNbgTt6JZQkLVjf35l+HvhoklOALwE/zeAHxq1JrgQeAd7e8xySpAXqVe5VdS8w23zPxX2OK0nqx79QlaQGWe6S1CDLXZIaZLlLUoMsd0lqkOUuSQ2y3CWpQZa7JDXIcpekBlnuktQgy12SGmS5S1KDLHdJapDlLkkNstwlqUGWuyQ1yHKXpAZZ7pLUIMtdkhpkuUtSgyx3SWqQ5S5JDbLcJalBlrskNchyl6QGWe6S1CDLXZIaZLlLUoMsd0lqkOUuSQ2y3CWpQb3LPclJST6f5OPd/XVJ7k6yv7td2z+mJGkhFuPK/Wpg39D97cCuqtoE7OruS5KWUa9yT3ImcAnwm0PDlwI7uuUdwGV9ziFJWri+V+7vB34R+PbQ2GRVPQHQ3X53z3NIkhYoVTXajsnbgLdW1b9MMg3826p6W5IjVbVmaLsnq+o58+5JtgHbACYnJ8/fuXPnSDkAjh49ysTExMj7rxRzLy9zz9/ex57qfYzJU+HQ04sQZhls3nD6s8vj9DjZsmXLnqqamm1dn3L/FeDdwDHgpcArgNuBHwKmq+qJJOuBmap69fMda2pqqnbv3j1SDoCZmRmmp6dH3n+lmHt5mXv+Nm6/q/cxrtl8jBv2nrwIaZbegesveXZ5nB4nSeYs95GnZarq2qo6s6o2ApcDn6qqdwF3Alu7zbYCd4x6DknSaJbide7XA29Ksh94U3dfkrSMFuV3pqqaAWa65f8LXLwYx5Ukjca/UJWkBlnuktQgy12SGmS5S1KDLHdJapDlLkkNstwlqUGWuyQ1yHKXpAZZ7pLUIMtdkhpkuUtSgyx3SWqQ5S5JDbLcJalBlrskNchyl6QGWe6S1CDLXZIaZLlLUoMsd0lq0MkrHUDS7DZuvwuAazYf44puWZovr9wlqUGWuyQ1yHKXpAZZ7pLUIMtdkhpkuUtSgyx3SWqQ5S5JDbLcJalBlrskNWjkck9yVpI/TLIvyf1Jru7G1yW5O8n+7nbt4sWVJM1Hnyv3Y8A1VfUPgQuBq5KcA2wHdlXVJmBXd1+StIxGLveqeqKqPtctfx3YB2wALgV2dJvtAC7rmVGStECpqv4HSTYCnwbOBR6pqjVD656squdMzSTZBmwDmJycPH/nzp0jn//o0aNMTEyMvP9KMffyGrfcex97CoDJU+HQ0yscZgTjlHvzhtOfXR6nx8mWLVv2VNXUbOt6l3uSCeCPgOuq6vYkR+ZT7sOmpqZq9+7dI2eYmZlhenp65P1XirmX17jlHn7L3xv2jt+7c49T7gPXX/Ls8jg9TpLMWe69Xi2T5MXA7wEfrarbu+FDSdZ369cDh/ucQ5K0cH1eLRPgZmBfVf3a0Ko7ga3d8lbgjtHjSZJG0ed3ptcD7wb2Jrm3G/sl4Hrg1iRXAo8Ab++VUJK0YCOXe1V9Bsgcqy8e9biSpP78C1VJapDlLkkNstwlqUGWuyQ1yHKXpAZZ7pLUIMtdkhpkuUtSgyx3SWqQ5S5JDbLcJalBlrskNWg83klfkpbQ8Q9GgcGHjFwxdH+pDX9QyGLyyl2SGmS5S1KDLHdJapDlLkkNstwlqUGWuyQ1yHKXpAZZ7pLUIMtdkhpkuUtSgyx3SWqQ5S5JDbLcJalBlrskNchyl6QGWe6S1CDLXZIaZLlLUoOa+Ji9vY89tawfi3XcUn08liT1tWRX7knekuTBJA8l2b5U55EkPdeSlHuSk4DfAH4cOAd4R5JzluJckqTnWqor99cBD1XVl6rqW8BO4NIlOpck6QSpqsU/aPLPgLdU1Xu6++8GLqiqnxvaZhuwrbv7auDBHqc8A/hqj/1XirmXl7mXl7mX3quq6pWzrViqJ1Qzy9h3/BSpqpuAmxblZMnuqppajGMtJ3MvL3MvL3OvrKWaljkInDV0/0zg8SU6lyTpBEtV7n8ObEpydpJTgMuBO5foXJKkEyzJtExVHUvyc8DvAycBH6qq+5fiXJ1Fmd5ZAeZeXuZeXuZeQUvyhKokaWX59gOS1CDLXZIaNNblvtre4iDJh5IcTnLf0Ni6JHcn2d/drh1ad22X/cEkbx4aPz/J3m7df0sy20tLFzP3WUn+MMm+JPcnuXocsid5aZLPJvlCl/t945B76JwnJfl8ko+PS+4kB7rz3Ztk9xjlXpPktiRf7B7nF41D7l6qaiy/GDxR+zDwvcApwBeAc1Y40xuA84D7hsb+M7C9W94O/Gq3fE6X+SXA2d33clK37rPARQz+XuATwI8vce71wHnd8suB/9PlW9XZu3NMdMsvBu4BLlztuYfy/xvgY8DHx+ixcgA444Sxcci9A3hPt3wKsGYccvf6nlc6QI//WRcBvz90/1rg2lWQayPfWe4PAuu75fXAg7PlZfDKoou6bb44NP4O4APL/D3cAbxpnLIDLwM+B1wwDrkZ/O3HLuCN/F25j0PuAzy33Fd1buAVwJfpXkAyLrn7fo3ztMwG4NGh+we7sdVmsqqeAOhuv7sbnyv/hm75xPFlkWQj8IMMroJXffZuauNe4DBwd1WNRW7g/cAvAt8eGhuH3AV8MsmeDN5CBFZ/7u8FvgJ8uJsG+80kp41B7l7Gudxf8C0OVrm58q/Y95VkAvg94F9V1V8/36azjK1I9qp6pqpey+BK+HVJzn2ezVdF7iRvAw5X1Z757jLL2Eo9Vl5fVecxeMfXq5K84Xm2XS25T2YwXfo/quoHgW8wmIaZy2rJ3cs4l/u4vMXBoSTrAbrbw934XPkPdssnji+pJC9mUOwfrarbu+GxyA5QVUeAGeAtrP7crwf+aZIDDN4x9Y1JPjIGuamqx7vbw8D/ZPAOsKs990HgYPdbHcBtDMp+tefuZZzLfVze4uBOYGu3vJXBfPbx8cuTvCTJ2cAm4LPdr4dfT3Jh90z8Tw3tsyS689wM7KuqXxuX7ElemWRNt3wq8KPAF1d77qq6tqrOrKqNDB63n6qqd6323ElOS/Ly48vAjwH3rfbcVfVXwKNJXt0NXQw8sNpz97bSk/59voC3Mnhlx8PAL6+CPLcATwB/y+Cn/JXAdzF44mx/d7tuaPtf7rI/yNCz7sAUg380DwM3csITQUuQ+x8z+PXyL4B7u6+3rvbswGuAz3e57wP+Qze+qnOf8D1M83dPqK7q3Azmrr/Qfd1//N/cas/dne+1wO7usfK/gLXjkLvPl28/IEkNGudpGUnSHCx3SWqQ5S5JDbLcJalBlrskNchyl6QGWe6S1KD/B6G2uXejWfc2AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# This 1D visualization of each numeric feature might inform your normalization\n",
    "# decisions.\n",
    "for feature_name in numeric_feature_names:\n",
    "  car_data.hist(column=feature_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "oiEpDZKSj8pN"
   },
   "source": [
    "###Train your model with numeric features + normalization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "c30Y6IiR8iVn",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "## Your code goes here\n",
    "# x_df.head()\n",
    "# y_series.head()\n",
    "x_df.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['symboling', 'normalized-losses', 'wheel-base', 'length', 'width', 'height', 'weight', 'engine-size', 'horsepower', 'peak-rpm', 'city-mpg', 'highway-mpg', 'bore', 'stroke', 'compression-ratio']\n"
     ]
    }
   ],
   "source": [
    "print(numeric_feature_names)\n",
    "\n",
    "model_feature_cols = [\n",
    "    tf.feature_column.numeric_column(feature_name, normalizer_fn=lambda val: val - x_df.mean()[feature_name] / x_df.std()[feature_name]) \n",
    "    for feature_name in numeric_feature_names\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Using default config.\n",
      "WARNING:tensorflow:Using temporary folder as model directory: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf\n",
      "INFO:tensorflow:Using config: {'_model_dir': '/var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true\n",
      "graph_options {\n",
      "  rewrite_options {\n",
      "    meta_optimizer_iterations: ONE\n",
      "  }\n",
      "}\n",
      ", '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...\n",
      "INFO:tensorflow:Saving checkpoints for 0 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...\n",
      "INFO:tensorflow:loss = 3190110500.0, step = 1\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 100...\n",
      "INFO:tensorflow:Saving checkpoints for 100 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 100...\n",
      "INFO:tensorflow:Loss for final step: 561099700.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:40:10Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-100\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.19521s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:40:11\n",
      "INFO:tensorflow:Saving dict for global step 100: average_loss = 52605596.0, global_step = 100, label/mean = 12949.43, loss = 829549800.0, prediction/mean = 12455.24\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 100: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-100\n",
      "scores {'average_loss': 52605596.0, 'label/mean': 12949.43, 'loss': 829549800.0, 'prediction/mean': 12455.24, 'global_step': 100}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-100\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 100...\n",
      "INFO:tensorflow:Saving checkpoints for 100 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 100...\n",
      "INFO:tensorflow:loss = 213835460.0, step = 101\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 200...\n",
      "INFO:tensorflow:Saving checkpoints for 200 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 200...\n",
      "INFO:tensorflow:Loss for final step: 685881660.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:40:12Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.21119s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:40:12\n",
      "INFO:tensorflow:Saving dict for global step 200: average_loss = 48679100.0, global_step = 200, label/mean = 12949.43, loss = 767632000.0, prediction/mean = 13060.464\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 200: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-200\n",
      "scores {'average_loss': 48679100.0, 'label/mean': 12949.43, 'loss': 767632000.0, 'prediction/mean': 13060.464, 'global_step': 200}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 200...\n",
      "INFO:tensorflow:Saving checkpoints for 200 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 200...\n",
      "INFO:tensorflow:loss = 448697900.0, step = 201\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 300...\n",
      "INFO:tensorflow:Saving checkpoints for 300 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 300...\n",
      "INFO:tensorflow:Loss for final step: 196279520.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:40:13Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-300\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.18156s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:40:14\n",
      "INFO:tensorflow:Saving dict for global step 300: average_loss = 45743100.0, global_step = 300, label/mean = 12949.43, loss = 721333500.0, prediction/mean = 13055.3125\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 300: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-300\n",
      "scores {'average_loss': 45743100.0, 'label/mean': 12949.43, 'loss': 721333500.0, 'prediction/mean': 13055.3125, 'global_step': 300}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-300\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 300...\n",
      "INFO:tensorflow:Saving checkpoints for 300 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 300...\n",
      "INFO:tensorflow:loss = 932427600.0, step = 301\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 400...\n",
      "INFO:tensorflow:Saving checkpoints for 400 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 400...\n",
      "INFO:tensorflow:Loss for final step: 401886750.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:40:15Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.20457s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:40:15\n",
      "INFO:tensorflow:Saving dict for global step 400: average_loss = 43209932.0, global_step = 400, label/mean = 12949.43, loss = 681387400.0, prediction/mean = 13115.827\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 400: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-400\n",
      "scores {'average_loss': 43209932.0, 'label/mean': 12949.43, 'loss': 681387400.0, 'prediction/mean': 13115.827, 'global_step': 400}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 400...\n",
      "INFO:tensorflow:Saving checkpoints for 400 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 400...\n",
      "INFO:tensorflow:loss = 463760000.0, step = 401\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 500...\n",
      "INFO:tensorflow:Saving checkpoints for 500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 500...\n",
      "INFO:tensorflow:Loss for final step: 1087146500.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:40:16Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.29255s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:40:17\n",
      "INFO:tensorflow:Saving dict for global step 500: average_loss = 41043620.0, global_step = 500, label/mean = 12949.43, loss = 647226300.0, prediction/mean = 12984.889\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 500: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-500\n",
      "scores {'average_loss': 41043620.0, 'label/mean': 12949.43, 'loss': 647226300.0, 'prediction/mean': 12984.889, 'global_step': 500}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 500...\n",
      "INFO:tensorflow:Saving checkpoints for 500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 500...\n",
      "INFO:tensorflow:loss = 153079650.0, step = 501\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 600...\n",
      "INFO:tensorflow:Saving checkpoints for 600 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 600...\n",
      "INFO:tensorflow:Loss for final step: 1067809400.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:40:18Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.19705s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:40:18\n",
      "INFO:tensorflow:Saving dict for global step 600: average_loss = 39119416.0, global_step = 600, label/mean = 12949.43, loss = 616883140.0, prediction/mean = 13038.657\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 600: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-600\n",
      "scores {'average_loss': 39119416.0, 'label/mean': 12949.43, 'loss': 616883140.0, 'prediction/mean': 13038.657, 'global_step': 600}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 600...\n",
      "INFO:tensorflow:Saving checkpoints for 600 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 600...\n",
      "INFO:tensorflow:loss = 477547330.0, step = 601\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 700...\n",
      "INFO:tensorflow:Saving checkpoints for 700 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 700...\n",
      "INFO:tensorflow:Loss for final step: 119583550.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:40:20Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-700\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.20103s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:40:20\n",
      "INFO:tensorflow:Saving dict for global step 700: average_loss = 37369692.0, global_step = 700, label/mean = 12949.43, loss = 589291260.0, prediction/mean = 13066.57\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 700: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-700\n",
      "scores {'average_loss': 37369692.0, 'label/mean': 12949.43, 'loss': 589291260.0, 'prediction/mean': 13066.57, 'global_step': 700}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-700\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 700...\n",
      "INFO:tensorflow:Saving checkpoints for 700 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 700...\n",
      "INFO:tensorflow:loss = 597961500.0, step = 701\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 800...\n",
      "INFO:tensorflow:Saving checkpoints for 800 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 800...\n",
      "INFO:tensorflow:Loss for final step: 429357570.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:40:21Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.18726s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:40:21\n",
      "INFO:tensorflow:Saving dict for global step 800: average_loss = 35940908.0, global_step = 800, label/mean = 12949.43, loss = 566760450.0, prediction/mean = 13022.142\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 800: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-800\n",
      "scores {'average_loss': 35940908.0, 'label/mean': 12949.43, 'loss': 566760450.0, 'prediction/mean': 13022.142, 'global_step': 800}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-800\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 800...\n",
      "INFO:tensorflow:Saving checkpoints for 800 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 800...\n",
      "INFO:tensorflow:loss = 448511420.0, step = 801\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 900...\n",
      "INFO:tensorflow:Saving checkpoints for 900 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 900...\n",
      "INFO:tensorflow:Loss for final step: 657667400.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:40:23Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-900\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.19594s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:40:23\n",
      "INFO:tensorflow:Saving dict for global step 900: average_loss = 34663970.0, global_step = 900, label/mean = 12949.43, loss = 546624100.0, prediction/mean = 12917.874\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 900: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-900\n",
      "scores {'average_loss': 34663970.0, 'label/mean': 12949.43, 'loss': 546624100.0, 'prediction/mean': 12917.874, 'global_step': 900}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-900\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 900...\n",
      "INFO:tensorflow:Saving checkpoints for 900 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 900...\n",
      "INFO:tensorflow:loss = 244503800.0, step = 901\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 1000...\n",
      "INFO:tensorflow:Saving checkpoints for 1000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 1000...\n",
      "INFO:tensorflow:Loss for final step: 426429200.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T13:40:24Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-1000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.17939s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-13:40:24\n",
      "INFO:tensorflow:Saving dict for global step 1000: average_loss = 33517558.0, global_step = 1000, label/mean = 12949.43, loss = 528546100.0, prediction/mean = 12786.005\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 1000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpd8w65_uf/model.ckpt-1000\n",
      "scores {'average_loss': 33517558.0, 'label/mean': 12949.43, 'loss': 528546100.0, 'prediction/mean': 12786.005, 'global_step': 1000}\n"
     ]
    }
   ],
   "source": [
    "# print('model features', model_feature_cols)\n",
    "\n",
    "train_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    batch_size=batch_size,\n",
    "    num_epochs=None,\n",
    "    shuffle=True)\n",
    "\n",
    "eval_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    batch_size=batch_size,\n",
    "    shuffle=False)\n",
    "\n",
    "est = tf.estimator.DNNRegressor(\n",
    "    feature_columns=model_feature_cols,\n",
    "    hidden_units=[64],\n",
    "    optimizer=tf.train.AdagradOptimizer(learning_rate=0.01)\n",
    ")\n",
    "\n",
    "num_print_statements=10\n",
    "num_train_steps=1000\n",
    "\n",
    "for _ in range(num_print_statements):\n",
    "    est.train(train_input_fn, steps=num_train_steps // num_print_statements)\n",
    "    scores = est.evaluate(eval_input_fn)\n",
    "    print('scores', scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "#@title Possible solution\n",
    "# This does Z-score normalization since the distributions for most features looked\n",
    "# roughly normally distributed.\n",
    "\n",
    "# Z-score normalization subtracts the mean and divides by the standard deviation,\n",
    "# to give a roughly standard normal distribution (mean = 0, std = 1) under a\n",
    "# normal distribution assumption. Epsilon prevents divide by zero.\n",
    "\n",
    "# With normalization, are you able to get the model working with\n",
    "# GradientDescentOptimizer? Z-score normalization doesn't seem to be able to get\n",
    "# SGD working. Maybe a different type of normalization would?\n",
    "\n",
    "batch_size = 16\n",
    "\n",
    "print(numeric_feature_names)\n",
    "x_df = car_data[numeric_feature_names]\n",
    "y_series = car_data['price']\n",
    "\n",
    "train_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    batch_size=batch_size,\n",
    "    num_epochs=None,\n",
    "    shuffle=True)\n",
    "\n",
    "eval_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    batch_size=batch_size,\n",
    "    shuffle=False)\n",
    "\n",
    "predict_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    batch_size=batch_size,\n",
    "    shuffle=False)\n",
    "\n",
    "# Epsilon prevents divide by zero.\n",
    "epsilon = 0.000001\n",
    "model_feature_columns = [\n",
    "    tf.feature_column.numeric_column(feature_name,\n",
    "                                     normalizer_fn=lambda val: (val - x_df.mean()[feature_name]) / (epsilon + x_df.std()[feature_name]))\n",
    "    for feature_name in numeric_feature_names\n",
    "]\n",
    "print('model_feature_columns', model_feature_columns)\n",
    "\n",
    "est = tf.estimator.DNNRegressor(\n",
    "    feature_columns=model_feature_columns,\n",
    "    hidden_units=[64],\n",
    "    optimizer=tf.train.AdagradOptimizer(learning_rate=0.01),\n",
    "  )\n",
    "\n",
    "# TRAIN\n",
    "num_print_statements = 10\n",
    "num_training_steps = 10000\n",
    "for _ in range(num_print_statements):\n",
    "  est.train(train_input_fn, steps=num_training_steps // num_print_statements)\n",
    "  scores = est.evaluate(eval_input_fn)\n",
    "  \n",
    "  # The `scores` dictionary has several metrics automatically generated by the \n",
    "  # canned Estimator.\n",
    "  # `average_loss` is the average loss for an individual example.\n",
    "  # `loss` is the summed loss for the batch.\n",
    "  # In addition to these scalar losses, you may find the visualization functions\n",
    "  # in the next cell helpful for debugging model quality.\n",
    "  print('scores', scores)\n",
    "\n",
    "scatter_plot_inference_grid(est, x_df, numeric_feature_names)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Task 3: Make your best model using only categorical features\n",
    "\n",
    "- Look at the possible feature columns for categorical features. They begin with `categorical_column_with_` in go/tf-ops.\n",
    "- You may find `dataframe[categorical_feature_names].unique()` helpful.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Using default config.\n",
      "WARNING:tensorflow:Using temporary folder as model directory: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo\n",
      "INFO:tensorflow:Using config: {'_model_dir': '/var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true\n",
      "graph_options {\n",
      "  rewrite_options {\n",
      "    meta_optimizer_iterations: ONE\n",
      "  }\n",
      "}\n",
      ", '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...\n",
      "INFO:tensorflow:Saving checkpoints for 0 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...\n",
      "INFO:tensorflow:loss = 2620492800.0, step = 1\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 100...\n",
      "INFO:tensorflow:Saving checkpoints for 100 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 100...\n",
      "INFO:tensorflow:Loss for final step: 3845584000.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:14:58Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-100\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.23973s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:14:58\n",
      "INFO:tensorflow:Saving dict for global step 100: average_loss = 231619630.0, global_step = 100, label/mean = 12949.43, loss = 3652463400.0, prediction/mean = 39.959244\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 100: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-100\n",
      "scores {'average_loss': 231619630.0, 'label/mean': 12949.43, 'loss': 3652463400.0, 'prediction/mean': 39.959244, 'global_step': 100}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-100\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 100...\n",
      "INFO:tensorflow:Saving checkpoints for 100 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 100...\n",
      "INFO:tensorflow:loss = 3505949700.0, step = 101\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 200...\n",
      "INFO:tensorflow:Saving checkpoints for 200 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 200...\n",
      "INFO:tensorflow:Loss for final step: 4361194500.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:00Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.22747s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:00\n",
      "INFO:tensorflow:Saving dict for global step 200: average_loss = 230653060.0, global_step = 200, label/mean = 12949.43, loss = 3637221000.0, prediction/mean = 77.9863\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 200: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-200\n",
      "scores {'average_loss': 230653060.0, 'label/mean': 12949.43, 'loss': 3637221000.0, 'prediction/mean': 77.9863, 'global_step': 200}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 200...\n",
      "INFO:tensorflow:Saving checkpoints for 200 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 200...\n",
      "INFO:tensorflow:loss = 2653252600.0, step = 201\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 300...\n",
      "INFO:tensorflow:Saving checkpoints for 300 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 300...\n",
      "INFO:tensorflow:Loss for final step: 1907127600.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:01Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-300\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.21749s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:01\n",
      "INFO:tensorflow:Saving dict for global step 300: average_loss = 229688450.0, global_step = 300, label/mean = 12949.43, loss = 3622010400.0, prediction/mean = 115.98481\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 300: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-300\n",
      "scores {'average_loss': 229688450.0, 'label/mean': 12949.43, 'loss': 3622010400.0, 'prediction/mean': 115.98481, 'global_step': 300}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-300\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 300...\n",
      "INFO:tensorflow:Saving checkpoints for 300 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 300...\n",
      "INFO:tensorflow:loss = 4088135400.0, step = 301\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 400...\n",
      "INFO:tensorflow:Saving checkpoints for 400 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 400...\n",
      "INFO:tensorflow:Loss for final step: 4261402400.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:03Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.22802s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:03\n",
      "INFO:tensorflow:Saving dict for global step 400: average_loss = 228742910.0, global_step = 400, label/mean = 12949.43, loss = 3607100000.0, prediction/mean = 153.35306\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 400: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-400\n",
      "scores {'average_loss': 228742910.0, 'label/mean': 12949.43, 'loss': 3607100000.0, 'prediction/mean': 153.35306, 'global_step': 400}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 400...\n",
      "INFO:tensorflow:Saving checkpoints for 400 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 400...\n",
      "INFO:tensorflow:loss = 6098113500.0, step = 401\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 500...\n",
      "INFO:tensorflow:Saving checkpoints for 500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 500...\n",
      "INFO:tensorflow:Loss for final step: 3922993200.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:05Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.22175s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:05\n",
      "INFO:tensorflow:Saving dict for global step 500: average_loss = 227806800.0, global_step = 500, label/mean = 12949.43, loss = 3592338000.0, prediction/mean = 190.50592\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 500: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-500\n",
      "scores {'average_loss': 227806800.0, 'label/mean': 12949.43, 'loss': 3592338000.0, 'prediction/mean': 190.50592, 'global_step': 500}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 500...\n",
      "INFO:tensorflow:Saving checkpoints for 500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 500...\n",
      "INFO:tensorflow:loss = 3085381600.0, step = 501\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 600...\n",
      "INFO:tensorflow:Saving checkpoints for 600 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 600...\n",
      "INFO:tensorflow:Loss for final step: 5098337300.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:06Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.20898s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:06\n",
      "INFO:tensorflow:Saving dict for global step 600: average_loss = 226866430.0, global_step = 600, label/mean = 12949.43, loss = 3577509000.0, prediction/mean = 227.8222\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 600: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-600\n",
      "scores {'average_loss': 226866430.0, 'label/mean': 12949.43, 'loss': 3577509000.0, 'prediction/mean': 227.8222, 'global_step': 600}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 600...\n",
      "INFO:tensorflow:Saving checkpoints for 600 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 600...\n",
      "INFO:tensorflow:loss = 3888822300.0, step = 601\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 700...\n",
      "INFO:tensorflow:Saving checkpoints for 700 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 700...\n",
      "INFO:tensorflow:Loss for final step: 5661876000.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:08Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-700\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.21948s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:08\n",
      "INFO:tensorflow:Saving dict for global step 700: average_loss = 225948430.0, global_step = 700, label/mean = 12949.43, loss = 3563032800.0, prediction/mean = 264.44043\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 700: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-700\n",
      "scores {'average_loss': 225948430.0, 'label/mean': 12949.43, 'loss': 3563032800.0, 'prediction/mean': 264.44043, 'global_step': 700}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-700\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 700...\n",
      "INFO:tensorflow:Saving checkpoints for 700 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 700...\n",
      "INFO:tensorflow:loss = 2827029000.0, step = 701\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 800...\n",
      "INFO:tensorflow:Saving checkpoints for 800 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 800...\n",
      "INFO:tensorflow:Loss for final step: 2275373000.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:09Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.21298s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:09\n",
      "INFO:tensorflow:Saving dict for global step 800: average_loss = 225037520.0, global_step = 800, label/mean = 12949.43, loss = 3548668400.0, prediction/mean = 300.84427\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 800: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-800\n",
      "scores {'average_loss': 225037520.0, 'label/mean': 12949.43, 'loss': 3548668400.0, 'prediction/mean': 300.84427, 'global_step': 800}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-800\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 800...\n",
      "INFO:tensorflow:Saving checkpoints for 800 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 800...\n",
      "INFO:tensorflow:loss = 2690331600.0, step = 801\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 900...\n",
      "INFO:tensorflow:Saving checkpoints for 900 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 900...\n",
      "INFO:tensorflow:Loss for final step: 3161626600.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:11Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-900\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.22566s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:11\n",
      "INFO:tensorflow:Saving dict for global step 900: average_loss = 224134740.0, global_step = 900, label/mean = 12949.43, loss = 3534432500.0, prediction/mean = 337.08292\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 900: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-900\n",
      "scores {'average_loss': 224134740.0, 'label/mean': 12949.43, 'loss': 3534432500.0, 'prediction/mean': 337.08292, 'global_step': 900}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-900\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 900...\n",
      "INFO:tensorflow:Saving checkpoints for 900 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 900...\n",
      "INFO:tensorflow:loss = 2681535000.0, step = 901\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 1000...\n",
      "INFO:tensorflow:Saving checkpoints for 1000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 1000...\n",
      "INFO:tensorflow:Loss for final step: 5873394000.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:12Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-1000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.23254s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:13\n",
      "INFO:tensorflow:Saving dict for global step 1000: average_loss = 223215540.0, global_step = 1000, label/mean = 12949.43, loss = 3519937300.0, prediction/mean = 373.95236\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 1000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpak6_6gwo/model.ckpt-1000\n",
      "scores {'average_loss': 223215540.0, 'label/mean': 12949.43, 'loss': 3519937300.0, 'prediction/mean': 373.95236, 'global_step': 1000}\n"
     ]
    }
   ],
   "source": [
    "\n",
    "# features dataframe\n",
    "x_categorical_df = car_data[categorical_feature_names]\n",
    "y_series = car_data[LABEL]\n",
    "batch_size = 16\n",
    "\n",
    "\n",
    "train_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_categorical_df,\n",
    "    y=y_series,\n",
    "    batch_size=batch_size,\n",
    "    shuffle=True,\n",
    "    num_epochs=None\n",
    ")\n",
    "\n",
    "eval_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_categorical_df,\n",
    "    y=y_series,\n",
    "    shuffle=False,\n",
    "    batch_size=batch_size\n",
    ")\n",
    "\n",
    "predict_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_categorical_df,\n",
    "    shuffle=False,\n",
    "    batch_size=batch_size\n",
    ")\n",
    "\n",
    "# feature columns\n",
    "model_categorical_cols = [\n",
    "    tf.feature_column.indicator_column(\n",
    "        tf.feature_column.categorical_column_with_vocabulary_list(feature_name, vocabulary_list=x_categorical_df[feature_name].unique())\n",
    "    ) for feature_name in categorical_feature_names\n",
    "]\n",
    "\n",
    "# Model\n",
    "est_categorical = tf.estimator.DNNRegressor(\n",
    "    feature_columns=model_categorical_cols,\n",
    "    hidden_units=[64],\n",
    "    optimizer=tf.train.AdagradOptimizer(learning_rate=0.01)\n",
    ")\n",
    "\n",
    "# train\n",
    "num_print_statements=10\n",
    "num_train_steps=1000\n",
    "\n",
    "for _ in range(num_print_statements):\n",
    "    est_categorical.train(train_input_fn, steps=num_train_steps // num_print_statements)\n",
    "    scores = est_categorical.evaluate(eval_input_fn)\n",
    "    print('scores', scores)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "cellView": "both",
    "colab": {},
    "colab_type": "code",
    "id": "RxYJy71zaZsy",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model_feature_columns [IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='body-style', vocabulary_list=('sedan', 'hatchback', 'wagon', 'hardtop', 'convertible'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='num-doors', vocabulary_list=('four', 'two', '?'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='fuel-system', vocabulary_list=('2bbl', '1bbl', 'mpfi', 'idi', '4bbl', 'spdi', 'spfi', 'mfi'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='drive-wheels', vocabulary_list=('fwd', 'rwd', '4wd'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='fuel-type', vocabulary_list=('gas', 'diesel'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='make', vocabulary_list=('mitsubishi', 'honda', 'nissan', 'audi', 'jaguar', 'chevrolet', 'mazda', 'volvo', 'porsche', 'subaru', 'toyota', 'mercedes-benz', 'volkswagen', 'plymouth', 'peugot', 'bmw', 'isuzu', 'saab', 'alfa-romero', 'dodge', 'mercury', 'renault'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='engine-location', vocabulary_list=('front', 'rear'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='engine-type', vocabulary_list=('ohc', 'ohcv', 'dohc', 'l', 'dohcv', 'ohcf', 'rotor'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='aspiration', vocabulary_list=('std', 'turbo'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='num-cylinders', vocabulary_list=('four', 'six', 'five', 'three', 'twelve', 'eight', 'two'), dtype=tf.string, default_value=-1, num_oov_buckets=0))]\n",
      "INFO:tensorflow:Using default config.\n",
      "WARNING:tensorflow:Using temporary folder as model directory: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv\n",
      "INFO:tensorflow:Using config: {'_model_dir': '/var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true\n",
      "graph_options {\n",
      "  rewrite_options {\n",
      "    meta_optimizer_iterations: ONE\n",
      "  }\n",
      "}\n",
      ", '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...\n",
      "INFO:tensorflow:Saving checkpoints for 0 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...\n",
      "INFO:tensorflow:loss = 2601490000.0, step = 1\n",
      "INFO:tensorflow:global_step/sec: 581.379\n",
      "INFO:tensorflow:loss = 2183082800.0, step = 101 (0.174 sec)\n",
      "INFO:tensorflow:global_step/sec: 776.85\n",
      "INFO:tensorflow:loss = 2886094300.0, step = 201 (0.128 sec)\n",
      "INFO:tensorflow:global_step/sec: 757.627\n",
      "INFO:tensorflow:loss = 2093586600.0, step = 301 (0.132 sec)\n",
      "INFO:tensorflow:global_step/sec: 764.813\n",
      "INFO:tensorflow:loss = 4561263000.0, step = 401 (0.131 sec)\n",
      "INFO:tensorflow:global_step/sec: 748.536\n",
      "INFO:tensorflow:loss = 4160828200.0, step = 501 (0.134 sec)\n",
      "INFO:tensorflow:global_step/sec: 769.432\n",
      "INFO:tensorflow:loss = 3059003400.0, step = 601 (0.130 sec)\n",
      "INFO:tensorflow:global_step/sec: 789.877\n",
      "INFO:tensorflow:loss = 2574578400.0, step = 701 (0.127 sec)\n",
      "INFO:tensorflow:global_step/sec: 758.369\n",
      "INFO:tensorflow:loss = 2401837300.0, step = 801 (0.132 sec)\n",
      "INFO:tensorflow:global_step/sec: 774.706\n",
      "INFO:tensorflow:loss = 3976916500.0, step = 901 (0.129 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 1000...\n",
      "INFO:tensorflow:Saving checkpoints for 1000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 1000...\n",
      "INFO:tensorflow:Loss for final step: 1357041900.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:50Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-1000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.20058s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:51\n",
      "INFO:tensorflow:Saving dict for global step 1000: average_loss = 225307920.0, global_step = 1000, label/mean = 12949.43, loss = 3552932400.0, prediction/mean = 290.17996\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 1000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-1000\n",
      "scores {'average_loss': 225307920.0, 'label/mean': 12949.43, 'loss': 3552932400.0, 'prediction/mean': 290.17996, 'global_step': 1000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-1000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 1000...\n",
      "INFO:tensorflow:Saving checkpoints for 1000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 1000...\n",
      "INFO:tensorflow:loss = 1960713500.0, step = 1001\n",
      "INFO:tensorflow:global_step/sec: 577.688\n",
      "INFO:tensorflow:loss = 3549198000.0, step = 1101 (0.174 sec)\n",
      "INFO:tensorflow:global_step/sec: 768.262\n",
      "INFO:tensorflow:loss = 4994021400.0, step = 1201 (0.130 sec)\n",
      "INFO:tensorflow:global_step/sec: 764.876\n",
      "INFO:tensorflow:loss = 3093034200.0, step = 1301 (0.131 sec)\n",
      "INFO:tensorflow:global_step/sec: 779.781\n",
      "INFO:tensorflow:loss = 2323083500.0, step = 1401 (0.128 sec)\n",
      "INFO:tensorflow:global_step/sec: 766.743\n",
      "INFO:tensorflow:loss = 2742912500.0, step = 1501 (0.130 sec)\n",
      "INFO:tensorflow:global_step/sec: 791.359\n",
      "INFO:tensorflow:loss = 1946997600.0, step = 1601 (0.126 sec)\n",
      "INFO:tensorflow:global_step/sec: 764.537\n",
      "INFO:tensorflow:loss = 4286107100.0, step = 1701 (0.131 sec)\n",
      "INFO:tensorflow:global_step/sec: 750.379\n",
      "INFO:tensorflow:loss = 5308849700.0, step = 1801 (0.133 sec)\n",
      "INFO:tensorflow:global_step/sec: 747.195\n",
      "INFO:tensorflow:loss = 4008755500.0, step = 1901 (0.134 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 2000...\n",
      "INFO:tensorflow:Saving checkpoints for 2000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 2000...\n",
      "INFO:tensorflow:Loss for final step: 4659482600.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:53Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-2000\n",
      "INFO:tensorflow:Running local_init_op.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.21081s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:53\n",
      "INFO:tensorflow:Saving dict for global step 2000: average_loss = 218266220.0, global_step = 2000, label/mean = 12949.43, loss = 3441890600.0, prediction/mean = 575.02997\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 2000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-2000\n",
      "scores {'average_loss': 218266220.0, 'label/mean': 12949.43, 'loss': 3441890600.0, 'prediction/mean': 575.02997, 'global_step': 2000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-2000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 2000...\n",
      "INFO:tensorflow:Saving checkpoints for 2000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 2000...\n",
      "INFO:tensorflow:loss = 3354153700.0, step = 2001\n",
      "INFO:tensorflow:global_step/sec: 477.659\n",
      "INFO:tensorflow:loss = 3490438100.0, step = 2101 (0.211 sec)\n",
      "INFO:tensorflow:global_step/sec: 618.465\n",
      "INFO:tensorflow:loss = 6571294700.0, step = 2201 (0.160 sec)\n",
      "INFO:tensorflow:global_step/sec: 615.324\n",
      "INFO:tensorflow:loss = 2777042000.0, step = 2301 (0.163 sec)\n",
      "INFO:tensorflow:global_step/sec: 620.055\n",
      "INFO:tensorflow:loss = 3429264100.0, step = 2401 (0.161 sec)\n",
      "INFO:tensorflow:global_step/sec: 630.179\n",
      "INFO:tensorflow:loss = 2027731500.0, step = 2501 (0.159 sec)\n",
      "INFO:tensorflow:global_step/sec: 620.779\n",
      "INFO:tensorflow:loss = 6872334300.0, step = 2601 (0.161 sec)\n",
      "INFO:tensorflow:global_step/sec: 720.493\n",
      "INFO:tensorflow:loss = 3591459000.0, step = 2701 (0.141 sec)\n",
      "INFO:tensorflow:global_step/sec: 619.049\n",
      "INFO:tensorflow:loss = 1395723000.0, step = 2801 (0.160 sec)\n",
      "INFO:tensorflow:global_step/sec: 670.804\n",
      "INFO:tensorflow:loss = 3851105800.0, step = 2901 (0.149 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 3000...\n",
      "INFO:tensorflow:Saving checkpoints for 3000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 3000...\n",
      "INFO:tensorflow:Loss for final step: 5976237600.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:56Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-3000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.21367s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:56\n",
      "INFO:tensorflow:Saving dict for global step 3000: average_loss = 211509710.0, global_step = 3000, label/mean = 12949.43, loss = 3335345400.0, prediction/mean = 854.6492\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 3000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-3000\n",
      "scores {'average_loss': 211509710.0, 'label/mean': 12949.43, 'loss': 3335345400.0, 'prediction/mean': 854.6492, 'global_step': 3000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-3000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 3000...\n",
      "INFO:tensorflow:Saving checkpoints for 3000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 3000...\n",
      "INFO:tensorflow:loss = 4058935300.0, step = 3001\n",
      "INFO:tensorflow:global_step/sec: 562.171\n",
      "INFO:tensorflow:loss = 2320575700.0, step = 3101 (0.179 sec)\n",
      "INFO:tensorflow:global_step/sec: 823.297\n",
      "INFO:tensorflow:loss = 5706108000.0, step = 3201 (0.121 sec)\n",
      "INFO:tensorflow:global_step/sec: 785.878\n",
      "INFO:tensorflow:loss = 3851297800.0, step = 3301 (0.127 sec)\n",
      "INFO:tensorflow:global_step/sec: 850.094\n",
      "INFO:tensorflow:loss = 6051324000.0, step = 3401 (0.118 sec)\n",
      "INFO:tensorflow:global_step/sec: 780.469\n",
      "INFO:tensorflow:loss = 3107396600.0, step = 3501 (0.128 sec)\n",
      "INFO:tensorflow:global_step/sec: 748.476\n",
      "INFO:tensorflow:loss = 3594987000.0, step = 3601 (0.134 sec)\n",
      "INFO:tensorflow:global_step/sec: 767.069\n",
      "INFO:tensorflow:loss = 3301454800.0, step = 3701 (0.130 sec)\n",
      "INFO:tensorflow:global_step/sec: 765.498\n",
      "INFO:tensorflow:loss = 3131428900.0, step = 3801 (0.131 sec)\n",
      "INFO:tensorflow:global_step/sec: 777.055\n",
      "INFO:tensorflow:loss = 2743719400.0, step = 3901 (0.129 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 4000...\n",
      "INFO:tensorflow:Saving checkpoints for 4000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 4000...\n",
      "INFO:tensorflow:Loss for final step: 2320268300.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:15:59Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-4000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.20203s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:15:59\n",
      "INFO:tensorflow:Saving dict for global step 4000: average_loss = 205001400.0, global_step = 4000, label/mean = 12949.43, loss = 3232714500.0, prediction/mean = 1130.181\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 4000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-4000\n",
      "scores {'average_loss': 205001400.0, 'label/mean': 12949.43, 'loss': 3232714500.0, 'prediction/mean': 1130.181, 'global_step': 4000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-4000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 4000...\n",
      "INFO:tensorflow:Saving checkpoints for 4000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 4000...\n",
      "INFO:tensorflow:loss = 3937693400.0, step = 4001\n",
      "INFO:tensorflow:global_step/sec: 576.296\n",
      "INFO:tensorflow:loss = 1637243600.0, step = 4101 (0.174 sec)\n",
      "INFO:tensorflow:global_step/sec: 765.557\n",
      "INFO:tensorflow:loss = 1685399400.0, step = 4201 (0.131 sec)\n",
      "INFO:tensorflow:global_step/sec: 758.702\n",
      "INFO:tensorflow:loss = 3295799300.0, step = 4301 (0.132 sec)\n",
      "INFO:tensorflow:global_step/sec: 762.379\n",
      "INFO:tensorflow:loss = 1151331600.0, step = 4401 (0.131 sec)\n",
      "INFO:tensorflow:global_step/sec: 770.833\n",
      "INFO:tensorflow:loss = 1832992400.0, step = 4501 (0.130 sec)\n",
      "INFO:tensorflow:global_step/sec: 770.184\n",
      "INFO:tensorflow:loss = 2225938700.0, step = 4601 (0.130 sec)\n",
      "INFO:tensorflow:global_step/sec: 763.791\n",
      "INFO:tensorflow:loss = 1578271500.0, step = 4701 (0.131 sec)\n",
      "INFO:tensorflow:global_step/sec: 764.947\n",
      "INFO:tensorflow:loss = 5511877600.0, step = 4801 (0.131 sec)\n",
      "INFO:tensorflow:global_step/sec: 767.537\n",
      "INFO:tensorflow:loss = 1902216000.0, step = 4901 (0.130 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 5000...\n",
      "INFO:tensorflow:Saving checkpoints for 5000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 5000...\n",
      "INFO:tensorflow:Loss for final step: 5894737000.0.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:16:01Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-5000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.19942s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:16:01\n",
      "INFO:tensorflow:Saving dict for global step 5000: average_loss = 198733800.0, global_step = 5000, label/mean = 12949.43, loss = 3133878800.0, prediction/mean = 1401.6932\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 5000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-5000\n",
      "scores {'average_loss': 198733800.0, 'label/mean': 12949.43, 'loss': 3133878800.0, 'prediction/mean': 1401.6932, 'global_step': 5000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-5000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 5000...\n",
      "INFO:tensorflow:Saving checkpoints for 5000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 5000...\n",
      "INFO:tensorflow:loss = 885828350.0, step = 5001\n",
      "INFO:tensorflow:global_step/sec: 576.782\n",
      "INFO:tensorflow:loss = 2459492900.0, step = 5101 (0.173 sec)\n",
      "INFO:tensorflow:global_step/sec: 801.903\n",
      "INFO:tensorflow:loss = 2430000400.0, step = 5201 (0.125 sec)\n",
      "INFO:tensorflow:global_step/sec: 793.494\n",
      "INFO:tensorflow:loss = 1359884800.0, step = 5301 (0.126 sec)\n",
      "INFO:tensorflow:global_step/sec: 796.928\n",
      "INFO:tensorflow:loss = 3001516300.0, step = 5401 (0.126 sec)\n",
      "INFO:tensorflow:global_step/sec: 790.115\n",
      "INFO:tensorflow:loss = 2268375600.0, step = 5501 (0.127 sec)\n",
      "INFO:tensorflow:global_step/sec: 809.656\n",
      "INFO:tensorflow:loss = 3525975000.0, step = 5601 (0.124 sec)\n",
      "INFO:tensorflow:global_step/sec: 775.326\n",
      "INFO:tensorflow:loss = 2891406300.0, step = 5701 (0.128 sec)\n",
      "INFO:tensorflow:global_step/sec: 745.029\n",
      "INFO:tensorflow:loss = 3089087000.0, step = 5801 (0.134 sec)\n",
      "INFO:tensorflow:global_step/sec: 802.099\n",
      "INFO:tensorflow:loss = 1486649600.0, step = 5901 (0.125 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 6000...\n",
      "INFO:tensorflow:Saving checkpoints for 6000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 6000...\n",
      "INFO:tensorflow:Loss for final step: 3867880700.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:16:04Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-6000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.20040s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:16:04\n",
      "INFO:tensorflow:Saving dict for global step 6000: average_loss = 192700450.0, global_step = 6000, label/mean = 12949.43, loss = 3038738000.0, prediction/mean = 1669.1442\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 6000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-6000\n",
      "scores {'average_loss': 192700450.0, 'label/mean': 12949.43, 'loss': 3038738000.0, 'prediction/mean': 1669.1442, 'global_step': 6000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-6000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 6000...\n",
      "INFO:tensorflow:Saving checkpoints for 6000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 6000...\n",
      "INFO:tensorflow:loss = 2229085700.0, step = 6001\n",
      "INFO:tensorflow:global_step/sec: 579.993\n",
      "INFO:tensorflow:loss = 1892653200.0, step = 6101 (0.173 sec)\n",
      "INFO:tensorflow:global_step/sec: 768.38\n",
      "INFO:tensorflow:loss = 1989144600.0, step = 6201 (0.130 sec)\n",
      "INFO:tensorflow:global_step/sec: 766.195\n",
      "INFO:tensorflow:loss = 3396996600.0, step = 6301 (0.130 sec)\n",
      "INFO:tensorflow:global_step/sec: 760.087\n",
      "INFO:tensorflow:loss = 3734897700.0, step = 6401 (0.132 sec)\n",
      "INFO:tensorflow:global_step/sec: 756.091\n",
      "INFO:tensorflow:loss = 1846556200.0, step = 6501 (0.132 sec)\n",
      "INFO:tensorflow:global_step/sec: 762.573\n",
      "INFO:tensorflow:loss = 4470868500.0, step = 6601 (0.131 sec)\n",
      "INFO:tensorflow:global_step/sec: 754.218\n",
      "INFO:tensorflow:loss = 4202920400.0, step = 6701 (0.133 sec)\n",
      "INFO:tensorflow:global_step/sec: 751.799\n",
      "INFO:tensorflow:loss = 4060660700.0, step = 6801 (0.133 sec)\n",
      "INFO:tensorflow:global_step/sec: 757.788\n",
      "INFO:tensorflow:loss = 2013129500.0, step = 6901 (0.132 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 7000...\n",
      "INFO:tensorflow:Saving checkpoints for 7000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 7000...\n",
      "INFO:tensorflow:Loss for final step: 2356639700.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:16:06Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-7000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.19982s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:16:07\n",
      "INFO:tensorflow:Saving dict for global step 7000: average_loss = 186888600.0, global_step = 7000, label/mean = 12949.43, loss = 2947089700.0, prediction/mean = 1932.8547\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 7000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-7000\n",
      "scores {'average_loss': 186888600.0, 'label/mean': 12949.43, 'loss': 2947089700.0, 'prediction/mean': 1932.8547, 'global_step': 7000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-7000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 7000...\n",
      "INFO:tensorflow:Saving checkpoints for 7000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 7000...\n",
      "INFO:tensorflow:loss = 6090047500.0, step = 7001\n",
      "INFO:tensorflow:global_step/sec: 571.164\n",
      "INFO:tensorflow:loss = 2099730200.0, step = 7101 (0.176 sec)\n",
      "INFO:tensorflow:global_step/sec: 810.013\n",
      "INFO:tensorflow:loss = 4027375600.0, step = 7201 (0.123 sec)\n",
      "INFO:tensorflow:global_step/sec: 817\n",
      "INFO:tensorflow:loss = 2803649000.0, step = 7301 (0.122 sec)\n",
      "INFO:tensorflow:global_step/sec: 776.815\n",
      "INFO:tensorflow:loss = 1527441500.0, step = 7401 (0.129 sec)\n",
      "INFO:tensorflow:global_step/sec: 792.309\n",
      "INFO:tensorflow:loss = 2302996500.0, step = 7501 (0.127 sec)\n",
      "INFO:tensorflow:global_step/sec: 782.462\n",
      "INFO:tensorflow:loss = 3585556700.0, step = 7601 (0.127 sec)\n",
      "INFO:tensorflow:global_step/sec: 797.727\n",
      "INFO:tensorflow:loss = 3141553200.0, step = 7701 (0.125 sec)\n",
      "INFO:tensorflow:global_step/sec: 803.639\n",
      "INFO:tensorflow:loss = 3987689500.0, step = 7801 (0.124 sec)\n",
      "INFO:tensorflow:global_step/sec: 691.076\n",
      "INFO:tensorflow:loss = 3810666000.0, step = 7901 (0.145 sec)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 8000...\n",
      "INFO:tensorflow:Saving checkpoints for 8000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 8000...\n",
      "INFO:tensorflow:Loss for final step: 2568008700.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:16:09Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-8000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.23169s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:16:09\n",
      "INFO:tensorflow:Saving dict for global step 8000: average_loss = 181296590.0, global_step = 8000, label/mean = 12949.43, loss = 2858908000.0, prediction/mean = 2192.623\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 8000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-8000\n",
      "scores {'average_loss': 181296590.0, 'label/mean': 12949.43, 'loss': 2858908000.0, 'prediction/mean': 2192.623, 'global_step': 8000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-8000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 8000...\n",
      "INFO:tensorflow:Saving checkpoints for 8000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 8000...\n",
      "INFO:tensorflow:loss = 1607637000.0, step = 8001\n",
      "INFO:tensorflow:global_step/sec: 468.81\n",
      "INFO:tensorflow:loss = 4269426200.0, step = 8101 (0.213 sec)\n",
      "INFO:tensorflow:global_step/sec: 601.912\n",
      "INFO:tensorflow:loss = 2948567600.0, step = 8201 (0.166 sec)\n",
      "INFO:tensorflow:global_step/sec: 669.86\n",
      "INFO:tensorflow:loss = 1502764800.0, step = 8301 (0.149 sec)\n",
      "INFO:tensorflow:global_step/sec: 622.913\n",
      "INFO:tensorflow:loss = 5937825000.0, step = 8401 (0.161 sec)\n",
      "INFO:tensorflow:global_step/sec: 608.743\n",
      "INFO:tensorflow:loss = 1052888100.0, step = 8501 (0.164 sec)\n",
      "INFO:tensorflow:global_step/sec: 642.686\n",
      "INFO:tensorflow:loss = 4052414200.0, step = 8601 (0.156 sec)\n",
      "INFO:tensorflow:global_step/sec: 624.91\n",
      "INFO:tensorflow:loss = 6956318700.0, step = 8701 (0.160 sec)\n",
      "INFO:tensorflow:global_step/sec: 587.158\n",
      "INFO:tensorflow:loss = 3378174700.0, step = 8801 (0.170 sec)\n",
      "INFO:tensorflow:global_step/sec: 606.086\n",
      "INFO:tensorflow:loss = 2379970600.0, step = 8901 (0.165 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 9000...\n",
      "INFO:tensorflow:Saving checkpoints for 9000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 9000...\n",
      "INFO:tensorflow:Loss for final step: 1476904600.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:16:12Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-9000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.23051s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:16:12\n",
      "INFO:tensorflow:Saving dict for global step 9000: average_loss = 175914620.0, global_step = 9000, label/mean = 12949.43, loss = 2774038300.0, prediction/mean = 2448.571\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 9000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-9000\n",
      "scores {'average_loss': 175914620.0, 'label/mean': 12949.43, 'loss': 2774038300.0, 'prediction/mean': 2448.571, 'global_step': 9000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-9000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 9000...\n",
      "INFO:tensorflow:Saving checkpoints for 9000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 9000...\n",
      "INFO:tensorflow:loss = 1406431200.0, step = 9001\n",
      "INFO:tensorflow:global_step/sec: 531.307\n",
      "INFO:tensorflow:loss = 1481766100.0, step = 9101 (0.189 sec)\n",
      "INFO:tensorflow:global_step/sec: 728.01\n",
      "INFO:tensorflow:loss = 3226531800.0, step = 9201 (0.137 sec)\n",
      "INFO:tensorflow:global_step/sec: 774.508\n",
      "INFO:tensorflow:loss = 2595458600.0, step = 9301 (0.129 sec)\n",
      "INFO:tensorflow:global_step/sec: 751.507\n",
      "INFO:tensorflow:loss = 1503659000.0, step = 9401 (0.133 sec)\n",
      "INFO:tensorflow:global_step/sec: 685.057\n",
      "INFO:tensorflow:loss = 2999050800.0, step = 9501 (0.146 sec)\n",
      "INFO:tensorflow:global_step/sec: 702.791\n",
      "INFO:tensorflow:loss = 1182470700.0, step = 9601 (0.142 sec)\n",
      "INFO:tensorflow:global_step/sec: 704.558\n",
      "INFO:tensorflow:loss = 3808241700.0, step = 9701 (0.142 sec)\n",
      "INFO:tensorflow:global_step/sec: 766.294\n",
      "INFO:tensorflow:loss = 3259755000.0, step = 9801 (0.130 sec)\n",
      "INFO:tensorflow:global_step/sec: 729.05\n",
      "INFO:tensorflow:loss = 4403901400.0, step = 9901 (0.137 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 10000...\n",
      "INFO:tensorflow:Saving checkpoints for 10000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 10000...\n",
      "INFO:tensorflow:Loss for final step: 2274791000.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:16:15Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-10000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.21695s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:16:15\n",
      "INFO:tensorflow:Saving dict for global step 10000: average_loss = 170736300.0, global_step = 10000, label/mean = 12949.43, loss = 2692380200.0, prediction/mean = 2700.7485\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 10000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmpo92_3ukv/model.ckpt-10000\n",
      "scores {'average_loss': 170736300.0, 'label/mean': 12949.43, 'loss': 2692380200.0, 'prediction/mean': 2700.7485, 'global_step': 10000}\n"
     ]
    }
   ],
   "source": [
    "#@title Possible solution\n",
    "# We have the full list of values that each feature takes on, and the list is\n",
    "# relatively small so we use categorical_column_with_vocabulary_list.\n",
    "\n",
    "batch_size = 16\n",
    "\n",
    "x_df = car_data[categorical_feature_names]\n",
    "y_series = car_data['price']\n",
    "\n",
    "train_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    batch_size=batch_size,\n",
    "    num_epochs=None,\n",
    "    shuffle=True)\n",
    "\n",
    "eval_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    batch_size=batch_size,\n",
    "    shuffle=False)\n",
    "\n",
    "predict_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    batch_size=batch_size,\n",
    "    shuffle=False)\n",
    "\n",
    "model_feature_columns = [\n",
    "    tf.feature_column.indicator_column(\n",
    "        tf.feature_column.categorical_column_with_vocabulary_list(\n",
    "            feature_name, vocabulary_list=car_data[feature_name].unique()))\n",
    "    for feature_name in categorical_feature_names\n",
    "]\n",
    "print('model_feature_columns', model_feature_columns)\n",
    "\n",
    "est = tf.estimator.DNNRegressor(\n",
    "    feature_columns=model_feature_columns,\n",
    "    hidden_units=[64],\n",
    "    optimizer=tf.train.AdagradOptimizer(learning_rate=0.01),\n",
    "  )\n",
    "\n",
    "# TRAIN\n",
    "num_print_statements = 10\n",
    "num_training_steps = 10000\n",
    "for _ in range(num_print_statements):\n",
    "  est.train(train_input_fn, steps=num_training_steps // num_print_statements)\n",
    "  scores = est.evaluate(eval_input_fn)\n",
    "  \n",
    "  # The `scores` dictionary has several metrics automatically generated by the\n",
    "  # canned Estimator.\n",
    "  # `average_loss` is the average loss for an individual example.\n",
    "  # `loss` is the summed loss for the batch.\n",
    "  # In addition to these scalar losses, you may find the visualization functions\n",
    "  # in the next cell helpful for debugging model quality.\n",
    "  print('scores', scores)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "gyBH5Wai_HTD"
   },
   "source": [
    "## Task 4: Using all the features, make the best model that you can make\n",
    "\n",
    "With all the features combined, your model should perform better than your earlier models using numerical and categorical models alone. Tune your model until that is the case."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "JNfCzC-q8edv",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>symboling</th>\n",
       "      <th>normalized-losses</th>\n",
       "      <th>wheel-base</th>\n",
       "      <th>length</th>\n",
       "      <th>width</th>\n",
       "      <th>height</th>\n",
       "      <th>weight</th>\n",
       "      <th>engine-size</th>\n",
       "      <th>horsepower</th>\n",
       "      <th>peak-rpm</th>\n",
       "      <th>...</th>\n",
       "      <th>body-style</th>\n",
       "      <th>num-doors</th>\n",
       "      <th>fuel-system</th>\n",
       "      <th>drive-wheels</th>\n",
       "      <th>fuel-type</th>\n",
       "      <th>make</th>\n",
       "      <th>engine-location</th>\n",
       "      <th>engine-type</th>\n",
       "      <th>aspiration</th>\n",
       "      <th>num-cylinders</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>86</th>\n",
       "      <td>1</td>\n",
       "      <td>125.00</td>\n",
       "      <td>96.30</td>\n",
       "      <td>172.40</td>\n",
       "      <td>65.40</td>\n",
       "      <td>51.60</td>\n",
       "      <td>2405</td>\n",
       "      <td>122</td>\n",
       "      <td>88.00</td>\n",
       "      <td>5000.00</td>\n",
       "      <td>...</td>\n",
       "      <td>sedan</td>\n",
       "      <td>four</td>\n",
       "      <td>2bbl</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>mitsubishi</td>\n",
       "      <td>front</td>\n",
       "      <td>ohc</td>\n",
       "      <td>std</td>\n",
       "      <td>four</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>76</th>\n",
       "      <td>2</td>\n",
       "      <td>161.00</td>\n",
       "      <td>93.70</td>\n",
       "      <td>157.30</td>\n",
       "      <td>64.40</td>\n",
       "      <td>50.80</td>\n",
       "      <td>1918</td>\n",
       "      <td>92</td>\n",
       "      <td>68.00</td>\n",
       "      <td>5500.00</td>\n",
       "      <td>...</td>\n",
       "      <td>hatchback</td>\n",
       "      <td>two</td>\n",
       "      <td>2bbl</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>mitsubishi</td>\n",
       "      <td>front</td>\n",
       "      <td>ohc</td>\n",
       "      <td>std</td>\n",
       "      <td>four</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>1</td>\n",
       "      <td>101.00</td>\n",
       "      <td>93.70</td>\n",
       "      <td>150.00</td>\n",
       "      <td>64.00</td>\n",
       "      <td>52.60</td>\n",
       "      <td>1956</td>\n",
       "      <td>92</td>\n",
       "      <td>76.00</td>\n",
       "      <td>6000.00</td>\n",
       "      <td>...</td>\n",
       "      <td>hatchback</td>\n",
       "      <td>two</td>\n",
       "      <td>1bbl</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>honda</td>\n",
       "      <td>front</td>\n",
       "      <td>ohc</td>\n",
       "      <td>std</td>\n",
       "      <td>four</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>103</th>\n",
       "      <td>0</td>\n",
       "      <td>108.00</td>\n",
       "      <td>100.40</td>\n",
       "      <td>184.60</td>\n",
       "      <td>66.50</td>\n",
       "      <td>55.10</td>\n",
       "      <td>3060</td>\n",
       "      <td>181</td>\n",
       "      <td>152.00</td>\n",
       "      <td>5200.00</td>\n",
       "      <td>...</td>\n",
       "      <td>sedan</td>\n",
       "      <td>four</td>\n",
       "      <td>mpfi</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>nissan</td>\n",
       "      <td>front</td>\n",
       "      <td>ohcv</td>\n",
       "      <td>std</td>\n",
       "      <td>six</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1</td>\n",
       "      <td>158.00</td>\n",
       "      <td>105.80</td>\n",
       "      <td>192.70</td>\n",
       "      <td>71.40</td>\n",
       "      <td>55.90</td>\n",
       "      <td>3086</td>\n",
       "      <td>131</td>\n",
       "      <td>140.00</td>\n",
       "      <td>5500.00</td>\n",
       "      <td>...</td>\n",
       "      <td>sedan</td>\n",
       "      <td>four</td>\n",
       "      <td>mpfi</td>\n",
       "      <td>fwd</td>\n",
       "      <td>gas</td>\n",
       "      <td>audi</td>\n",
       "      <td>front</td>\n",
       "      <td>ohc</td>\n",
       "      <td>turbo</td>\n",
       "      <td>five</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 25 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     symboling  normalized-losses  wheel-base  length  width  height  weight  \\\n",
       "86           1             125.00       96.30  172.40  65.40   51.60    2405   \n",
       "76           2             161.00       93.70  157.30  64.40   50.80    1918   \n",
       "34           1             101.00       93.70  150.00  64.00   52.60    1956   \n",
       "103          0             108.00      100.40  184.60  66.50   55.10    3060   \n",
       "8            1             158.00      105.80  192.70  71.40   55.90    3086   \n",
       "\n",
       "     engine-size  horsepower  peak-rpm  ...  body-style  num-doors  \\\n",
       "86           122       88.00   5000.00  ...       sedan       four   \n",
       "76            92       68.00   5500.00  ...   hatchback        two   \n",
       "34            92       76.00   6000.00  ...   hatchback        two   \n",
       "103          181      152.00   5200.00  ...       sedan       four   \n",
       "8            131      140.00   5500.00  ...       sedan       four   \n",
       "\n",
       "     fuel-system  drive-wheels  fuel-type        make engine-location  \\\n",
       "86          2bbl           fwd        gas  mitsubishi           front   \n",
       "76          2bbl           fwd        gas  mitsubishi           front   \n",
       "34          1bbl           fwd        gas       honda           front   \n",
       "103         mpfi           fwd        gas      nissan           front   \n",
       "8           mpfi           fwd        gas        audi           front   \n",
       "\n",
       "    engine-type aspiration num-cylinders  \n",
       "86          ohc        std          four  \n",
       "76          ohc        std          four  \n",
       "34          ohc        std          four  \n",
       "103        ohcv        std           six  \n",
       "8           ohc      turbo          five  \n",
       "\n",
       "[5 rows x 25 columns]"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch_size=16\n",
    "\n",
    "x_df = car_data[numeric_feature_names + categorical_feature_names]\n",
    "y_series = car_data[LABEL]\n",
    "\n",
    "x_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# model functions\n",
    "train_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    shuffle=True,\n",
    "    batch_size=batch_size,\n",
    "    num_epochs=None\n",
    ")\n",
    "\n",
    "eval_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    shuffle=False,\n",
    "    batch_size=batch_size\n",
    ")\n",
    "\n",
    "predict_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    shuffle=False,\n",
    "    batch_size=batch_size\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "all features cols [NumericColumn(key='symboling', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f8f641ca0>), NumericColumn(key='normalized-losses', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f8e98b940>), NumericColumn(key='wheel-base', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f8e98b310>), NumericColumn(key='length', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f8f5dedc0>), NumericColumn(key='width', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f984bee50>), NumericColumn(key='height', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f984be310>), NumericColumn(key='weight', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f922c6040>), NumericColumn(key='engine-size', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f922c6160>), NumericColumn(key='horsepower', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f922c61f0>), NumericColumn(key='peak-rpm', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f922c6280>), NumericColumn(key='city-mpg', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f922c6310>), NumericColumn(key='highway-mpg', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f922c63a0>), NumericColumn(key='bore', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f922c6430>), NumericColumn(key='stroke', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f922c64c0>), NumericColumn(key='compression-ratio', shape=(1,), default_value=None, dtype=tf.float32, normalizer_fn=<function <listcomp>.<lambda> at 0x7f8f922c6550>), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='body-style', vocabulary_list=('sedan', 'hatchback', 'wagon', 'hardtop', 'convertible'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='num-doors', vocabulary_list=('four', 'two', '?'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='fuel-system', vocabulary_list=('2bbl', '1bbl', 'mpfi', 'idi', '4bbl', 'spdi', 'spfi', 'mfi'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='drive-wheels', vocabulary_list=('fwd', 'rwd', '4wd'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='fuel-type', vocabulary_list=('gas', 'diesel'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='make', vocabulary_list=('mitsubishi', 'honda', 'nissan', 'audi', 'jaguar', 'chevrolet', 'mazda', 'volvo', 'porsche', 'subaru', 'toyota', 'mercedes-benz', 'volkswagen', 'plymouth', 'peugot', 'bmw', 'isuzu', 'saab', 'alfa-romero', 'dodge', 'mercury', 'renault'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='engine-location', vocabulary_list=('front', 'rear'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='engine-type', vocabulary_list=('ohc', 'ohcv', 'dohc', 'l', 'dohcv', 'ohcf', 'rotor'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='aspiration', vocabulary_list=('std', 'turbo'), dtype=tf.string, default_value=-1, num_oov_buckets=0)), IndicatorColumn(categorical_column=VocabularyListCategoricalColumn(key='num-cylinders', vocabulary_list=('four', 'six', 'five', 'three', 'twelve', 'eight', 'two'), dtype=tf.string, default_value=-1, num_oov_buckets=0))]\n"
     ]
    }
   ],
   "source": [
    "# feature columns\n",
    "epsilon = 0.00001\n",
    "\n",
    "numeric_feature_cols = [\n",
    "    tf.feature_column.numeric_column(feature_name, normalizer_fn=lambda val: (val-x_df[feature_name].mean())/(epsilon +x_df[feature_name].std()))\n",
    "    for feature_name in numeric_feature_names\n",
    "]\n",
    "\n",
    "categorical_feature_cols = [\n",
    "    tf.feature_column.indicator_column(tf.feature_column.categorical_column_with_vocabulary_list(feature_col, vocabulary_list=x_df[feature_col].unique()))\n",
    "    for feature_col in categorical_feature_names\n",
    "]\n",
    "\n",
    "all_features = numeric_feature_cols + categorical_feature_cols\n",
    "\n",
    "print('all features cols', all_features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Using default config.\n",
      "WARNING:tensorflow:Using temporary folder as model directory: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34\n",
      "INFO:tensorflow:Using config: {'_model_dir': '/var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true\n",
      "graph_options {\n",
      "  rewrite_options {\n",
      "    meta_optimizer_iterations: ONE\n",
      "  }\n",
      "}\n",
      ", '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_service': None, '_cluster_spec': ClusterSpec({}), '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}\n"
     ]
    }
   ],
   "source": [
    "model = tf.estimator.DNNRegressor(\n",
    "    feature_columns=all_features,\n",
    "    hidden_units=[64],\n",
    "    optimizer=tf.train.AdagradOptimizer(learning_rate=0.01)\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 0...\n",
      "INFO:tensorflow:Saving checkpoints for 0 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 0...\n",
      "INFO:tensorflow:loss = 2509623600.0, step = 1\n",
      "INFO:tensorflow:global_step/sec: 388.345\n",
      "INFO:tensorflow:loss = 859243600.0, step = 101 (0.259 sec)\n",
      "INFO:tensorflow:global_step/sec: 470.482\n",
      "INFO:tensorflow:loss = 2157260000.0, step = 201 (0.214 sec)\n",
      "INFO:tensorflow:global_step/sec: 515.886\n",
      "INFO:tensorflow:loss = 451967900.0, step = 301 (0.192 sec)\n",
      "INFO:tensorflow:global_step/sec: 483.962\n",
      "INFO:tensorflow:loss = 1304288000.0, step = 401 (0.209 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 500...\n",
      "INFO:tensorflow:Saving checkpoints for 500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 500...\n",
      "INFO:tensorflow:Loss for final step: 481300100.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:46:43Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.29194s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:46:43\n",
      "INFO:tensorflow:Saving dict for global step 500: average_loss = 53614696.0, global_step = 500, label/mean = 12949.43, loss = 845462500.0, prediction/mean = 12270.376\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 500: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-500\n",
      "scores {'average_loss': 53614696.0, 'label/mean': 12949.43, 'loss': 845462500.0, 'prediction/mean': 12270.376, 'global_step': 500}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 500...\n",
      "INFO:tensorflow:Saving checkpoints for 500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 500...\n",
      "INFO:tensorflow:loss = 226901100.0, step = 501\n",
      "INFO:tensorflow:global_step/sec: 305.382\n",
      "INFO:tensorflow:loss = 1195455100.0, step = 601 (0.330 sec)\n",
      "INFO:tensorflow:global_step/sec: 497.964\n",
      "INFO:tensorflow:loss = 1300890400.0, step = 701 (0.200 sec)\n",
      "INFO:tensorflow:global_step/sec: 501.94\n",
      "INFO:tensorflow:loss = 208225900.0, step = 801 (0.204 sec)\n",
      "INFO:tensorflow:global_step/sec: 454.326\n",
      "INFO:tensorflow:loss = 832018600.0, step = 901 (0.216 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 1000...\n",
      "INFO:tensorflow:Saving checkpoints for 1000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 1000...\n",
      "INFO:tensorflow:Loss for final step: 223244720.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:46:46Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-1000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.25200s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:46:46\n",
      "INFO:tensorflow:Saving dict for global step 1000: average_loss = 49352096.0, global_step = 1000, label/mean = 12949.43, loss = 778244540.0, prediction/mean = 13275.867\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 1000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-1000\n",
      "scores {'average_loss': 49352096.0, 'label/mean': 12949.43, 'loss': 778244540.0, 'prediction/mean': 13275.867, 'global_step': 1000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-1000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 1000...\n",
      "INFO:tensorflow:Saving checkpoints for 1000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 1000...\n",
      "INFO:tensorflow:loss = 621138370.0, step = 1001\n",
      "INFO:tensorflow:global_step/sec: 326.157\n",
      "INFO:tensorflow:loss = 566273700.0, step = 1101 (0.312 sec)\n",
      "INFO:tensorflow:global_step/sec: 420.007\n",
      "INFO:tensorflow:loss = 1798593000.0, step = 1201 (0.237 sec)\n",
      "INFO:tensorflow:global_step/sec: 461.608\n",
      "INFO:tensorflow:loss = 1307807500.0, step = 1301 (0.218 sec)\n",
      "INFO:tensorflow:global_step/sec: 428.255\n",
      "INFO:tensorflow:loss = 557911500.0, step = 1401 (0.229 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 1500...\n",
      "INFO:tensorflow:Saving checkpoints for 1500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 1500...\n",
      "INFO:tensorflow:Loss for final step: 400174300.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:46:48Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-1500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.25831s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:46:49\n",
      "INFO:tensorflow:Saving dict for global step 1500: average_loss = 46719796.0, global_step = 1500, label/mean = 12949.43, loss = 736735200.0, prediction/mean = 13322.607\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 1500: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-1500\n",
      "scores {'average_loss': 46719796.0, 'label/mean': 12949.43, 'loss': 736735200.0, 'prediction/mean': 13322.607, 'global_step': 1500}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-1500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 1500...\n",
      "INFO:tensorflow:Saving checkpoints for 1500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 1500...\n",
      "INFO:tensorflow:loss = 1170052900.0, step = 1501\n",
      "INFO:tensorflow:global_step/sec: 378.05\n",
      "INFO:tensorflow:loss = 1621321500.0, step = 1601 (0.268 sec)\n",
      "INFO:tensorflow:global_step/sec: 422.942\n",
      "INFO:tensorflow:loss = 1425545000.0, step = 1701 (0.234 sec)\n",
      "INFO:tensorflow:global_step/sec: 469.243\n",
      "INFO:tensorflow:loss = 608635900.0, step = 1801 (0.217 sec)\n",
      "INFO:tensorflow:global_step/sec: 523.227\n",
      "INFO:tensorflow:loss = 378842400.0, step = 1901 (0.190 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 2000...\n",
      "INFO:tensorflow:Saving checkpoints for 2000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 2000...\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Loss for final step: 533901820.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:46:51Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-2000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.24509s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:46:51\n",
      "INFO:tensorflow:Saving dict for global step 2000: average_loss = 44403164.0, global_step = 2000, label/mean = 12949.43, loss = 700203700.0, prediction/mean = 13183.762\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 2000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-2000\n",
      "scores {'average_loss': 44403164.0, 'label/mean': 12949.43, 'loss': 700203700.0, 'prediction/mean': 13183.762, 'global_step': 2000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-2000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 2000...\n",
      "INFO:tensorflow:Saving checkpoints for 2000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 2000...\n",
      "INFO:tensorflow:loss = 382418880.0, step = 2001\n",
      "INFO:tensorflow:global_step/sec: 399.14\n",
      "INFO:tensorflow:loss = 795415940.0, step = 2101 (0.253 sec)\n",
      "INFO:tensorflow:global_step/sec: 540.291\n",
      "INFO:tensorflow:loss = 902794750.0, step = 2201 (0.185 sec)\n",
      "INFO:tensorflow:global_step/sec: 475.346\n",
      "INFO:tensorflow:loss = 457910240.0, step = 2301 (0.209 sec)\n",
      "INFO:tensorflow:global_step/sec: 554.146\n",
      "INFO:tensorflow:loss = 548406460.0, step = 2401 (0.182 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 2500...\n",
      "INFO:tensorflow:Saving checkpoints for 2500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 2500...\n",
      "INFO:tensorflow:Loss for final step: 235496830.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:46:54Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-2500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.26280s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:46:54\n",
      "INFO:tensorflow:Saving dict for global step 2500: average_loss = 42182440.0, global_step = 2500, label/mean = 12949.43, loss = 665184640.0, prediction/mean = 13274.789\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 2500: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-2500\n",
      "scores {'average_loss': 42182440.0, 'label/mean': 12949.43, 'loss': 665184640.0, 'prediction/mean': 13274.789, 'global_step': 2500}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-2500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 2500...\n",
      "INFO:tensorflow:Saving checkpoints for 2500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 2500...\n",
      "INFO:tensorflow:loss = 387092800.0, step = 2501\n",
      "INFO:tensorflow:global_step/sec: 345.717\n",
      "INFO:tensorflow:loss = 486785440.0, step = 2601 (0.292 sec)\n",
      "INFO:tensorflow:global_step/sec: 431.36\n",
      "INFO:tensorflow:loss = 909453300.0, step = 2701 (0.233 sec)\n",
      "INFO:tensorflow:global_step/sec: 423.574\n",
      "INFO:tensorflow:loss = 2281684000.0, step = 2801 (0.235 sec)\n",
      "INFO:tensorflow:global_step/sec: 469.808\n",
      "INFO:tensorflow:loss = 540214340.0, step = 2901 (0.213 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 3000...\n",
      "INFO:tensorflow:Saving checkpoints for 3000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 3000...\n",
      "INFO:tensorflow:Loss for final step: 454669200.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:46:57Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-3000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.27732s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:46:57\n",
      "INFO:tensorflow:Saving dict for global step 3000: average_loss = 40096720.0, global_step = 3000, label/mean = 12949.43, loss = 632294400.0, prediction/mean = 13301.427\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 3000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-3000\n",
      "scores {'average_loss': 40096720.0, 'label/mean': 12949.43, 'loss': 632294400.0, 'prediction/mean': 13301.427, 'global_step': 3000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-3000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 3000...\n",
      "INFO:tensorflow:Saving checkpoints for 3000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 3000...\n",
      "INFO:tensorflow:loss = 233290320.0, step = 3001\n",
      "INFO:tensorflow:global_step/sec: 367.303\n",
      "INFO:tensorflow:loss = 953151600.0, step = 3101 (0.273 sec)\n",
      "INFO:tensorflow:global_step/sec: 489.282\n",
      "INFO:tensorflow:loss = 1271814100.0, step = 3201 (0.204 sec)\n",
      "INFO:tensorflow:global_step/sec: 457.259\n",
      "INFO:tensorflow:loss = 427206660.0, step = 3301 (0.219 sec)\n",
      "INFO:tensorflow:global_step/sec: 439.005\n",
      "INFO:tensorflow:loss = 1041127550.0, step = 3401 (0.228 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 3500...\n",
      "INFO:tensorflow:Saving checkpoints for 3500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 3500...\n",
      "INFO:tensorflow:Loss for final step: 137757300.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:47:00Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-3500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.26905s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:47:00\n",
      "INFO:tensorflow:Saving dict for global step 3500: average_loss = 38212364.0, global_step = 3500, label/mean = 12949.43, loss = 602579600.0, prediction/mean = 13303.996\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 3500: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-3500\n",
      "scores {'average_loss': 38212364.0, 'label/mean': 12949.43, 'loss': 602579600.0, 'prediction/mean': 13303.996, 'global_step': 3500}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-3500\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 3500...\n",
      "INFO:tensorflow:Saving checkpoints for 3500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 3500...\n",
      "INFO:tensorflow:loss = 178807920.0, step = 3501\n",
      "INFO:tensorflow:global_step/sec: 351.714\n",
      "INFO:tensorflow:loss = 1341966600.0, step = 3601 (0.285 sec)\n",
      "INFO:tensorflow:global_step/sec: 434.054\n",
      "INFO:tensorflow:loss = 233130620.0, step = 3701 (0.233 sec)\n",
      "INFO:tensorflow:global_step/sec: 441.472\n",
      "INFO:tensorflow:loss = 496774000.0, step = 3801 (0.223 sec)\n",
      "INFO:tensorflow:global_step/sec: 429.646\n",
      "INFO:tensorflow:loss = 486248500.0, step = 3901 (0.234 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 4000...\n",
      "INFO:tensorflow:Saving checkpoints for 4000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 4000...\n",
      "INFO:tensorflow:Loss for final step: 1122100000.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:47:03Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-4000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.26629s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:47:03\n",
      "INFO:tensorflow:Saving dict for global step 4000: average_loss = 36515410.0, global_step = 4000, label/mean = 12949.43, loss = 575819900.0, prediction/mean = 13202.362\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 4000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-4000\n",
      "scores {'average_loss': 36515410.0, 'label/mean': 12949.43, 'loss': 575819900.0, 'prediction/mean': 13202.362, 'global_step': 4000}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-4000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 4000...\n",
      "INFO:tensorflow:Saving checkpoints for 4000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 4000...\n",
      "INFO:tensorflow:loss = 154802750.0, step = 4001\n",
      "INFO:tensorflow:global_step/sec: 298.116\n",
      "INFO:tensorflow:loss = 447911500.0, step = 4101 (0.338 sec)\n",
      "INFO:tensorflow:global_step/sec: 422.567\n",
      "INFO:tensorflow:loss = 491527260.0, step = 4201 (0.237 sec)\n",
      "INFO:tensorflow:global_step/sec: 476.39\n",
      "INFO:tensorflow:loss = 958283600.0, step = 4301 (0.209 sec)\n",
      "INFO:tensorflow:global_step/sec: 494.594\n",
      "INFO:tensorflow:loss = 212676260.0, step = 4401 (0.202 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 4500...\n",
      "INFO:tensorflow:Saving checkpoints for 4500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 4500...\n",
      "INFO:tensorflow:Loss for final step: 295418560.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:47:06Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-4500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.27969s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:47:06\n",
      "INFO:tensorflow:Saving dict for global step 4500: average_loss = 34997108.0, global_step = 4500, label/mean = 12949.43, loss = 551877500.0, prediction/mean = 13185.173\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 4500: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-4500\n",
      "scores {'average_loss': 34997108.0, 'label/mean': 12949.43, 'loss': 551877500.0, 'prediction/mean': 13185.173, 'global_step': 4500}\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-4500\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 4500...\n",
      "INFO:tensorflow:Saving checkpoints for 4500 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 4500...\n",
      "INFO:tensorflow:loss = 694851650.0, step = 4501\n",
      "INFO:tensorflow:global_step/sec: 340.932\n",
      "INFO:tensorflow:loss = 104953140.0, step = 4601 (0.296 sec)\n",
      "INFO:tensorflow:global_step/sec: 429.442\n",
      "INFO:tensorflow:loss = 178425380.0, step = 4701 (0.233 sec)\n",
      "INFO:tensorflow:global_step/sec: 421.955\n",
      "INFO:tensorflow:loss = 675800060.0, step = 4801 (0.236 sec)\n",
      "INFO:tensorflow:global_step/sec: 424.293\n",
      "INFO:tensorflow:loss = 1193190700.0, step = 4901 (0.235 sec)\n",
      "INFO:tensorflow:Calling checkpoint listeners before saving checkpoint 5000...\n",
      "INFO:tensorflow:Saving checkpoints for 5000 into /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt.\n",
      "INFO:tensorflow:Calling checkpoint listeners after saving checkpoint 5000...\n",
      "INFO:tensorflow:Loss for final step: 874489700.0.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-08-30T15:47:09Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-5000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.25417s\n",
      "INFO:tensorflow:Finished evaluation at 2020-08-30-15:47:09\n",
      "INFO:tensorflow:Saving dict for global step 5000: average_loss = 33634644.0, global_step = 5000, label/mean = 12949.43, loss = 530392450.0, prediction/mean = 13212.309\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 5000: /var/folders/7l/1w7jpt51333brnwb5g8pcg1w0000gn/T/tmp10sypz34/model.ckpt-5000\n",
      "scores {'average_loss': 33634644.0, 'label/mean': 12949.43, 'loss': 530392450.0, 'prediction/mean': 13212.309, 'global_step': 5000}\n"
     ]
    }
   ],
   "source": [
    "num_print_statements=10\n",
    "num_train_steps=5000\n",
    "\n",
    "for _ in range(num_print_statements):\n",
    "    model.train(train_input_fn, steps=num_train_steps // num_print_statements)\n",
    "    scores = model.evaluate(eval_input_fn)\n",
    "    print('scores', scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "results = model.predict(predict_input_fn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<generator object Estimator.predict at 0x7f8f99f00820>\n"
     ]
    }
   ],
   "source": [
    "print(results)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "cellView": "both",
    "colab": {},
    "colab_type": "code",
    "id": "CmaGHWFVGKMr",
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "#@title Possible solution\n",
    "# This is a first pass at a model that uses all the features.\n",
    "# Do you have any improvements?\n",
    "\n",
    "batch_size = 16\n",
    "\n",
    "x_df = car_data[numeric_feature_names + categorical_feature_names]\n",
    "y_series = car_data['price']\n",
    "\n",
    "train_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    batch_size=batch_size,\n",
    "    num_epochs=None,\n",
    "    shuffle=True)\n",
    "\n",
    "eval_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    y=y_series,\n",
    "    batch_size=batch_size,\n",
    "    shuffle=False)\n",
    "\n",
    "predict_input_fn = tf.estimator.inputs.pandas_input_fn(\n",
    "    x=x_df,\n",
    "    batch_size=batch_size,\n",
    "    shuffle=False)\n",
    "\n",
    "epsilon = 0.000001\n",
    "model_feature_columns = [\n",
    "    tf.feature_column.indicator_column(\n",
    "        tf.feature_column.categorical_column_with_vocabulary_list(\n",
    "            feature_name, vocabulary_list=car_data[feature_name].unique()))\n",
    "    for feature_name in categorical_feature_names\n",
    "] + [\n",
    "    tf.feature_column.numeric_column(feature_name,\n",
    "                                     normalizer_fn=lambda val: (val - x_df.mean()[feature_name]) / (epsilon + x_df.std()[feature_name]))\n",
    "    for feature_name in numeric_feature_names\n",
    "]\n",
    "\n",
    "\n",
    "print('model_feature_columns', model_feature_columns)\n",
    "\n",
    "est = tf.estimator.DNNRegressor(\n",
    "    feature_columns=model_feature_columns,\n",
    "    hidden_units=[64],\n",
    "    optimizer=tf.train.AdagradOptimizer(learning_rate=0.01),\n",
    "  )\n",
    "\n",
    "# TRAIN\n",
    "num_print_statements = 10\n",
    "num_training_steps = 10000\n",
    "for _ in range(num_print_statements):\n",
    "  est.train(train_input_fn, steps=num_training_steps // num_print_statements)\n",
    "  scores = est.evaluate(eval_input_fn)\n",
    "  \n",
    "  # The `scores` dictionary has several metrics automatically generated by the \n",
    "  # canned Estimator.\n",
    "  # `average_loss` is the average loss for an individual example.\n",
    "  # `loss` is the summed loss for the batch.\n",
    "  # In addition to these scalar losses, you may find the visualization functions\n",
    "  # in the next cell helpful for debugging model quality.\n",
    "  print('scores', scores)\n",
    "\n"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "intro_to_modeling.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}