Neural Networks.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "from collections import namedtuple\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from sklearn.metrics import explained_variance_score, \\\n",
    "    mean_absolute_error, \\\n",
    "    median_absolute_error\n",
    "from sklearn.model_selection import train_test_split\n",
    "import csv\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "variables = [\"date\", \"meantempm\", \"meandewptm\", \"meanpressurem\", \"maxhumidity\", \"minhumidity\", \"maxtempm\",\n",
    "            \"mintempm\", \"maxdewptm\", \"mindewptm\", \"maxpressurem\", \"minpressurem\", \"precipm\"]\n",
    "\n",
    "records = []\n",
    "\n",
    "AnnualWeatherReport = namedtuple(\"AnnualWeatherReport\", variables)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\users\\naman\\appdata\\local\\programs\\python\\python37\\lib\\site-packages\\ipykernel_launcher.py:54: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n"
     ]
    }
   ],
   "source": [
    "# read in the csv data into a pandas data frame and set the date as the index\n",
    "with open(\"ottawa_raw_data.txt\", \"r\") as raw_data:\n",
    "    r = csv.reader(raw_data, delimiter=\",\")\n",
    "    next(r)\n",
    "    for row in r:\n",
    "        records.append(AnnualWeatherReport(\n",
    "            date=row[0],\n",
    "            meantempm=row[1],\n",
    "            meandewptm=row[2],\n",
    "            meanpressurem=row[3],\n",
    "            maxhumidity=row[4],\n",
    "            minhumidity=row[5],\n",
    "            maxtempm=row[6],\n",
    "            mintempm=row[7],\n",
    "            maxdewptm=row[8],\n",
    "            mindewptm=row[9],\n",
    "            maxpressurem=row[10],\n",
    "            minpressurem=row[11],\n",
    "            precipm=row[12]\n",
    "        ))\n",
    "\n",
    "\n",
    "df = pd.DataFrame(records, columns=variables).set_index('date')\n",
    "\n",
    "def add_nth_day_prior_features(df, target_variable, days_prior):\n",
    "    rows = df.shape[0]\n",
    "    nth_prior_values = [None]*days_prior + [df[target_variable][i-days_prior] for i in range(days_prior, rows)]\n",
    "    col_name = \"{}_{}\".format(target_variable, days_prior)\n",
    "    df[col_name] = nth_prior_values\n",
    "\n",
    "\n",
    "for target_variable in variables:\n",
    "    if target_variable != 'date':\n",
    "        for i in range(1, 4):\n",
    "            add_nth_day_prior_features(df, target_variable, i)\n",
    "\n",
    "\n",
    "variables_removed = [variable for variable in variables if variable not in ['meantempm', 'mintempm', 'maxtempm']]\n",
    "variables_maintained = [col for col in df.columns if col not in variables_removed]\n",
    "\n",
    "df = df[variables_maintained]\n",
    "\n",
    "df = df.apply(pd.to_numeric, errors='coerce')\n",
    "# print(df.info())\n",
    "\n",
    "spread = df.describe().T\n",
    "IQR = spread['75%'] - spread['25%']\n",
    "spread['outliers'] = (spread['min']<(spread['25%']-(3*IQR)))|(spread['max'] > (spread['75%']+3*IQR))\n",
    "\n",
    "# If Needed\n",
    "for precip_col in ['precipm_1', 'precipm_2', 'precipm_3']:\n",
    "    # create a boolean array of values representing nans\n",
    "    missing_vals = pd.isnull(df[precip_col])\n",
    "    df[precip_col][missing_vals] = 0\n",
    "\n",
    "df = df.dropna()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "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>count</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>min</th>\n",
       "      <th>25%</th>\n",
       "      <th>50%</th>\n",
       "      <th>75%</th>\n",
       "      <th>max</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>meantempm</th>\n",
       "      <td>149.0</td>\n",
       "      <td>41.993289</td>\n",
       "      <td>17.741194</td>\n",
       "      <td>3.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>39.0</td>\n",
       "      <td>54.0</td>\n",
       "      <td>95.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxtempm</th>\n",
       "      <td>149.0</td>\n",
       "      <td>15.080537</td>\n",
       "      <td>14.471555</td>\n",
       "      <td>-20.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>16.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>61.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mintempm</th>\n",
       "      <td>149.0</td>\n",
       "      <td>72.040268</td>\n",
       "      <td>30.336375</td>\n",
       "      <td>9.0</td>\n",
       "      <td>64.0</td>\n",
       "      <td>85.0</td>\n",
       "      <td>93.0</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>meantempm_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>41.899329</td>\n",
       "      <td>17.708887</td>\n",
       "      <td>3.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>39.0</td>\n",
       "      <td>52.0</td>\n",
       "      <td>95.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>meantempm_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>41.704698</td>\n",
       "      <td>17.578974</td>\n",
       "      <td>3.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>39.0</td>\n",
       "      <td>52.0</td>\n",
       "      <td>95.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>meantempm_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>41.389262</td>\n",
       "      <td>17.288377</td>\n",
       "      <td>3.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>39.0</td>\n",
       "      <td>52.0</td>\n",
       "      <td>95.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>meandewptm_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>34.053691</td>\n",
       "      <td>16.226424</td>\n",
       "      <td>-4.4</td>\n",
       "      <td>22.7</td>\n",
       "      <td>33.9</td>\n",
       "      <td>42.9</td>\n",
       "      <td>80.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>meandewptm_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>33.891275</td>\n",
       "      <td>16.124046</td>\n",
       "      <td>-4.4</td>\n",
       "      <td>22.7</td>\n",
       "      <td>33.4</td>\n",
       "      <td>42.8</td>\n",
       "      <td>80.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>meandewptm_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>33.621477</td>\n",
       "      <td>15.827467</td>\n",
       "      <td>-4.4</td>\n",
       "      <td>22.7</td>\n",
       "      <td>33.2</td>\n",
       "      <td>42.4</td>\n",
       "      <td>80.9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>meanpressurem_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>25.543624</td>\n",
       "      <td>15.675546</td>\n",
       "      <td>-15.0</td>\n",
       "      <td>14.0</td>\n",
       "      <td>27.0</td>\n",
       "      <td>34.0</td>\n",
       "      <td>66.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>meanpressurem_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>25.402685</td>\n",
       "      <td>15.584521</td>\n",
       "      <td>-15.0</td>\n",
       "      <td>14.0</td>\n",
       "      <td>27.0</td>\n",
       "      <td>34.0</td>\n",
       "      <td>66.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>meanpressurem_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>25.174497</td>\n",
       "      <td>15.261120</td>\n",
       "      <td>-15.0</td>\n",
       "      <td>14.0</td>\n",
       "      <td>27.0</td>\n",
       "      <td>34.0</td>\n",
       "      <td>66.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxhumidity_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>29.335570</td>\n",
       "      <td>14.636871</td>\n",
       "      <td>-9.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>28.0</td>\n",
       "      <td>37.0</td>\n",
       "      <td>70.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxhumidity_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>29.127517</td>\n",
       "      <td>14.373064</td>\n",
       "      <td>-9.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>28.0</td>\n",
       "      <td>37.0</td>\n",
       "      <td>70.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxhumidity_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>28.885906</td>\n",
       "      <td>14.047954</td>\n",
       "      <td>-9.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>28.0</td>\n",
       "      <td>36.0</td>\n",
       "      <td>70.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>minhumidity_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>22.712081</td>\n",
       "      <td>14.460921</td>\n",
       "      <td>-15.4</td>\n",
       "      <td>14.8</td>\n",
       "      <td>22.5</td>\n",
       "      <td>30.4</td>\n",
       "      <td>65.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>minhumidity_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>22.553020</td>\n",
       "      <td>14.284690</td>\n",
       "      <td>-15.4</td>\n",
       "      <td>14.8</td>\n",
       "      <td>22.5</td>\n",
       "      <td>29.9</td>\n",
       "      <td>65.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>minhumidity_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>22.314765</td>\n",
       "      <td>13.879016</td>\n",
       "      <td>-15.4</td>\n",
       "      <td>14.8</td>\n",
       "      <td>22.5</td>\n",
       "      <td>29.8</td>\n",
       "      <td>65.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxtempm_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>15.114094</td>\n",
       "      <td>14.500480</td>\n",
       "      <td>-20.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>16.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>61.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxtempm_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>14.993289</td>\n",
       "      <td>14.374572</td>\n",
       "      <td>-20.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>16.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>61.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxtempm_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>14.765101</td>\n",
       "      <td>13.901385</td>\n",
       "      <td>-20.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>16.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>61.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mintempm_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>72.489933</td>\n",
       "      <td>30.055044</td>\n",
       "      <td>9.0</td>\n",
       "      <td>64.0</td>\n",
       "      <td>85.0</td>\n",
       "      <td>93.0</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mintempm_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>72.953020</td>\n",
       "      <td>29.833623</td>\n",
       "      <td>9.0</td>\n",
       "      <td>64.0</td>\n",
       "      <td>85.0</td>\n",
       "      <td>93.0</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mintempm_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>73.489933</td>\n",
       "      <td>29.593100</td>\n",
       "      <td>9.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>85.0</td>\n",
       "      <td>93.0</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxdewptm_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>43.362416</td>\n",
       "      <td>25.738607</td>\n",
       "      <td>0.0</td>\n",
       "      <td>24.0</td>\n",
       "      <td>49.0</td>\n",
       "      <td>63.0</td>\n",
       "      <td>93.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxdewptm_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>43.758389</td>\n",
       "      <td>25.613460</td>\n",
       "      <td>0.0</td>\n",
       "      <td>25.0</td>\n",
       "      <td>50.0</td>\n",
       "      <td>63.0</td>\n",
       "      <td>93.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxdewptm_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>44.261745</td>\n",
       "      <td>25.583566</td>\n",
       "      <td>0.0</td>\n",
       "      <td>25.0</td>\n",
       "      <td>52.0</td>\n",
       "      <td>64.0</td>\n",
       "      <td>93.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mindewptm_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>29.715436</td>\n",
       "      <td>0.257778</td>\n",
       "      <td>29.1</td>\n",
       "      <td>29.5</td>\n",
       "      <td>29.7</td>\n",
       "      <td>29.9</td>\n",
       "      <td>30.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mindewptm_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>29.713423</td>\n",
       "      <td>0.259849</td>\n",
       "      <td>29.1</td>\n",
       "      <td>29.5</td>\n",
       "      <td>29.7</td>\n",
       "      <td>29.9</td>\n",
       "      <td>30.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mindewptm_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>29.712752</td>\n",
       "      <td>0.260791</td>\n",
       "      <td>29.1</td>\n",
       "      <td>29.5</td>\n",
       "      <td>29.7</td>\n",
       "      <td>29.9</td>\n",
       "      <td>30.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxpressurem_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>29.582550</td>\n",
       "      <td>0.275762</td>\n",
       "      <td>28.9</td>\n",
       "      <td>29.4</td>\n",
       "      <td>29.6</td>\n",
       "      <td>29.7</td>\n",
       "      <td>30.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxpressurem_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>29.581879</td>\n",
       "      <td>0.276331</td>\n",
       "      <td>28.9</td>\n",
       "      <td>29.4</td>\n",
       "      <td>29.6</td>\n",
       "      <td>29.7</td>\n",
       "      <td>30.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>maxpressurem_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>29.580537</td>\n",
       "      <td>0.278189</td>\n",
       "      <td>28.9</td>\n",
       "      <td>29.4</td>\n",
       "      <td>29.6</td>\n",
       "      <td>29.7</td>\n",
       "      <td>30.3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>minpressurem_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>29.451007</td>\n",
       "      <td>0.282273</td>\n",
       "      <td>28.8</td>\n",
       "      <td>29.3</td>\n",
       "      <td>29.5</td>\n",
       "      <td>29.6</td>\n",
       "      <td>30.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>minpressurem_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>29.450336</td>\n",
       "      <td>0.282753</td>\n",
       "      <td>28.8</td>\n",
       "      <td>29.3</td>\n",
       "      <td>29.5</td>\n",
       "      <td>29.6</td>\n",
       "      <td>30.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>minpressurem_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>29.448993</td>\n",
       "      <td>0.284419</td>\n",
       "      <td>28.8</td>\n",
       "      <td>29.3</td>\n",
       "      <td>29.5</td>\n",
       "      <td>29.6</td>\n",
       "      <td>30.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>precipm_1</th>\n",
       "      <td>149.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>precipm_2</th>\n",
       "      <td>149.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>precipm_3</th>\n",
       "      <td>149.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 count       mean        std   min   25%   50%   75%    max\n",
       "meantempm        149.0  41.993289  17.741194   3.0  30.0  39.0  54.0   95.0\n",
       "maxtempm         149.0  15.080537  14.471555 -20.0   7.0  16.0  23.0   61.0\n",
       "mintempm         149.0  72.040268  30.336375   9.0  64.0  85.0  93.0  100.0\n",
       "meantempm_1      149.0  41.899329  17.708887   3.0  30.0  39.0  52.0   95.0\n",
       "meantempm_2      149.0  41.704698  17.578974   3.0  30.0  39.0  52.0   95.0\n",
       "meantempm_3      149.0  41.389262  17.288377   3.0  30.0  39.0  52.0   95.0\n",
       "meandewptm_1     149.0  34.053691  16.226424  -4.4  22.7  33.9  42.9   80.9\n",
       "meandewptm_2     149.0  33.891275  16.124046  -4.4  22.7  33.4  42.8   80.9\n",
       "meandewptm_3     149.0  33.621477  15.827467  -4.4  22.7  33.2  42.4   80.9\n",
       "meanpressurem_1  149.0  25.543624  15.675546 -15.0  14.0  27.0  34.0   66.0\n",
       "meanpressurem_2  149.0  25.402685  15.584521 -15.0  14.0  27.0  34.0   66.0\n",
       "meanpressurem_3  149.0  25.174497  15.261120 -15.0  14.0  27.0  34.0   66.0\n",
       "maxhumidity_1    149.0  29.335570  14.636871  -9.0  23.0  28.0  37.0   70.0\n",
       "maxhumidity_2    149.0  29.127517  14.373064  -9.0  23.0  28.0  37.0   70.0\n",
       "maxhumidity_3    149.0  28.885906  14.047954  -9.0  23.0  28.0  36.0   70.0\n",
       "minhumidity_1    149.0  22.712081  14.460921 -15.4  14.8  22.5  30.4   65.0\n",
       "minhumidity_2    149.0  22.553020  14.284690 -15.4  14.8  22.5  29.9   65.0\n",
       "minhumidity_3    149.0  22.314765  13.879016 -15.4  14.8  22.5  29.8   65.0\n",
       "maxtempm_1       149.0  15.114094  14.500480 -20.0   7.0  16.0  23.0   61.0\n",
       "maxtempm_2       149.0  14.993289  14.374572 -20.0   7.0  16.0  23.0   61.0\n",
       "maxtempm_3       149.0  14.765101  13.901385 -20.0   7.0  16.0  23.0   61.0\n",
       "mintempm_1       149.0  72.489933  30.055044   9.0  64.0  85.0  93.0  100.0\n",
       "mintempm_2       149.0  72.953020  29.833623   9.0  64.0  85.0  93.0  100.0\n",
       "mintempm_3       149.0  73.489933  29.593100   9.0  65.0  85.0  93.0  100.0\n",
       "maxdewptm_1      149.0  43.362416  25.738607   0.0  24.0  49.0  63.0   93.0\n",
       "maxdewptm_2      149.0  43.758389  25.613460   0.0  25.0  50.0  63.0   93.0\n",
       "maxdewptm_3      149.0  44.261745  25.583566   0.0  25.0  52.0  64.0   93.0\n",
       "mindewptm_1      149.0  29.715436   0.257778  29.1  29.5  29.7  29.9   30.4\n",
       "mindewptm_2      149.0  29.713423   0.259849  29.1  29.5  29.7  29.9   30.4\n",
       "mindewptm_3      149.0  29.712752   0.260791  29.1  29.5  29.7  29.9   30.4\n",
       "maxpressurem_1   149.0  29.582550   0.275762  28.9  29.4  29.6  29.7   30.3\n",
       "maxpressurem_2   149.0  29.581879   0.276331  28.9  29.4  29.6  29.7   30.3\n",
       "maxpressurem_3   149.0  29.580537   0.278189  28.9  29.4  29.6  29.7   30.3\n",
       "minpressurem_1   149.0  29.451007   0.282273  28.8  29.3  29.5  29.6   30.2\n",
       "minpressurem_2   149.0  29.450336   0.282753  28.8  29.3  29.5  29.6   30.2\n",
       "minpressurem_3   149.0  29.448993   0.284419  28.8  29.3  29.5  29.6   30.2\n",
       "precipm_1        149.0   0.000000   0.000000   0.0   0.0   0.0   0.0    0.0\n",
       "precipm_2        149.0   0.000000   0.000000   0.0   0.0   0.0   0.0    0.0\n",
       "precipm_3        149.0   0.000000   0.000000   0.0   0.0   0.0   0.0    0.0"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe().T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Index: 149 entries, 04/01/2020 to 31/05/2020\n",
      "Data columns (total 39 columns):\n",
      " #   Column           Non-Null Count  Dtype  \n",
      "---  ------           --------------  -----  \n",
      " 0   meantempm        149 non-null    int64  \n",
      " 1   maxtempm         149 non-null    int64  \n",
      " 2   mintempm         149 non-null    int64  \n",
      " 3   meantempm_1      149 non-null    float64\n",
      " 4   meantempm_2      149 non-null    float64\n",
      " 5   meantempm_3      149 non-null    float64\n",
      " 6   meandewptm_1     149 non-null    float64\n",
      " 7   meandewptm_2     149 non-null    float64\n",
      " 8   meandewptm_3     149 non-null    float64\n",
      " 9   meanpressurem_1  149 non-null    float64\n",
      " 10  meanpressurem_2  149 non-null    float64\n",
      " 11  meanpressurem_3  149 non-null    float64\n",
      " 12  maxhumidity_1    149 non-null    float64\n",
      " 13  maxhumidity_2    149 non-null    float64\n",
      " 14  maxhumidity_3    149 non-null    float64\n",
      " 15  minhumidity_1    149 non-null    float64\n",
      " 16  minhumidity_2    149 non-null    float64\n",
      " 17  minhumidity_3    149 non-null    float64\n",
      " 18  maxtempm_1       149 non-null    float64\n",
      " 19  maxtempm_2       149 non-null    float64\n",
      " 20  maxtempm_3       149 non-null    float64\n",
      " 21  mintempm_1       149 non-null    float64\n",
      " 22  mintempm_2       149 non-null    float64\n",
      " 23  mintempm_3       149 non-null    float64\n",
      " 24  maxdewptm_1      149 non-null    float64\n",
      " 25  maxdewptm_2      149 non-null    float64\n",
      " 26  maxdewptm_3      149 non-null    float64\n",
      " 27  mindewptm_1      149 non-null    float64\n",
      " 28  mindewptm_2      149 non-null    float64\n",
      " 29  mindewptm_3      149 non-null    float64\n",
      " 30  maxpressurem_1   149 non-null    float64\n",
      " 31  maxpressurem_2   149 non-null    float64\n",
      " 32  maxpressurem_3   149 non-null    float64\n",
      " 33  minpressurem_1   149 non-null    float64\n",
      " 34  minpressurem_2   149 non-null    float64\n",
      " 35  minpressurem_3   149 non-null    float64\n",
      " 36  precipm_1        149 non-null    float64\n",
      " 37  precipm_2        149 non-null    float64\n",
      " 38  precipm_3        149 non-null    float64\n",
      "dtypes: float64(36), int64(3)\n",
      "memory usage: 46.6+ KB\n"
     ]
    }
   ],
   "source": [
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.drop(['mintempm', 'maxtempm'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Index: 149 entries, 04/01/2020 to 31/05/2020\n",
      "Data columns (total 37 columns):\n",
      " #   Column           Non-Null Count  Dtype  \n",
      "---  ------           --------------  -----  \n",
      " 0   meantempm        149 non-null    int64  \n",
      " 1   meantempm_1      149 non-null    float64\n",
      " 2   meantempm_2      149 non-null    float64\n",
      " 3   meantempm_3      149 non-null    float64\n",
      " 4   meandewptm_1     149 non-null    float64\n",
      " 5   meandewptm_2     149 non-null    float64\n",
      " 6   meandewptm_3     149 non-null    float64\n",
      " 7   meanpressurem_1  149 non-null    float64\n",
      " 8   meanpressurem_2  149 non-null    float64\n",
      " 9   meanpressurem_3  149 non-null    float64\n",
      " 10  maxhumidity_1    149 non-null    float64\n",
      " 11  maxhumidity_2    149 non-null    float64\n",
      " 12  maxhumidity_3    149 non-null    float64\n",
      " 13  minhumidity_1    149 non-null    float64\n",
      " 14  minhumidity_2    149 non-null    float64\n",
      " 15  minhumidity_3    149 non-null    float64\n",
      " 16  maxtempm_1       149 non-null    float64\n",
      " 17  maxtempm_2       149 non-null    float64\n",
      " 18  maxtempm_3       149 non-null    float64\n",
      " 19  mintempm_1       149 non-null    float64\n",
      " 20  mintempm_2       149 non-null    float64\n",
      " 21  mintempm_3       149 non-null    float64\n",
      " 22  maxdewptm_1      149 non-null    float64\n",
      " 23  maxdewptm_2      149 non-null    float64\n",
      " 24  maxdewptm_3      149 non-null    float64\n",
      " 25  mindewptm_1      149 non-null    float64\n",
      " 26  mindewptm_2      149 non-null    float64\n",
      " 27  mindewptm_3      149 non-null    float64\n",
      " 28  maxpressurem_1   149 non-null    float64\n",
      " 29  maxpressurem_2   149 non-null    float64\n",
      " 30  maxpressurem_3   149 non-null    float64\n",
      " 31  minpressurem_1   149 non-null    float64\n",
      " 32  minpressurem_2   149 non-null    float64\n",
      " 33  minpressurem_3   149 non-null    float64\n",
      " 34  precipm_1        149 non-null    float64\n",
      " 35  precipm_2        149 non-null    float64\n",
      " 36  precipm_3        149 non-null    float64\n",
      "dtypes: float64(36), int64(1)\n",
      "memory usage: 44.2+ KB\n"
     ]
    }
   ],
   "source": [
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df[[col for col in df.columns if col != 'meantempm']]\n",
    "# y will be a pandas series of the meantempm\n",
    "y = df['meantempm']\n",
    "\n",
    "# split data into training set and a temporary set using sklearn.model_selection.traing_test_split\n",
    "X_train, X_tmp, y_train, y_tmp = train_test_split(X, y, test_size=0.2, random_state=23)\n",
    "\n",
    "# take the remaining 20% of data in X_tmp, y_tmp and split them evenly\n",
    "X_test, X_val, y_test, y_val = train_test_split(X_tmp, y_tmp, test_size=0.5, random_state=23)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Using default config.\n",
      "INFO:tensorflow:Using config: {'_model_dir': 'tf_wx_model', '_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": [
    "feature_cols = [tf.feature_column.numeric_column(col) for col in X.columns]\n",
    "regressor = tf.estimator.DNNRegressor(feature_columns=feature_cols,\n",
    "                                      hidden_units=[50, 50],\n",
    "                                      model_dir='tf_wx_model')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-36400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 36400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 53.340256, step = 36400\n",
      "INFO:tensorflow:global_step/sec: 77.2917\n",
      "INFO:tensorflow:loss = 42.146183, step = 36500 (1.294 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 36600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 52.290016.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:15Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-36600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 1.11129s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:16\n",
      "INFO:tensorflow:Saving dict for global step 36600: average_loss = 99.79377, global_step = 36600, label/mean = 47.333332, loss = 99.79377, prediction/mean = 45.038815\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 36600: tf_wx_model\\model.ckpt-36600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:17Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-36600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.81884s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:18\n",
      "INFO:tensorflow:Saving dict for global step 36600: average_loss = 99.79377, global_step = 36600, label/mean = 47.333332, loss = 99.79377, prediction/mean = 45.038815\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 36600: tf_wx_model\\model.ckpt-36600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-36600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 36600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 44.730274, step = 36600\n",
      "INFO:tensorflow:global_step/sec: 121.319\n",
      "INFO:tensorflow:loss = 34.82431, step = 36700 (0.826 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 36800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 61.204117.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:23Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-36800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.83666s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:24\n",
      "INFO:tensorflow:Saving dict for global step 36800: average_loss = 99.942154, global_step = 36800, label/mean = 47.333332, loss = 99.942154, prediction/mean = 45.02842\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 36800: tf_wx_model\\model.ckpt-36800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:25Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-36800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.72363s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:26\n",
      "INFO:tensorflow:Saving dict for global step 36800: average_loss = 99.942154, global_step = 36800, label/mean = 47.333332, loss = 99.942154, prediction/mean = 45.02842\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 36800: tf_wx_model\\model.ckpt-36800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-36800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 36800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 59.516125, step = 36800\n",
      "INFO:tensorflow:global_step/sec: 121.43\n",
      "INFO:tensorflow:loss = 63.233887, step = 36900 (0.841 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 37000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 39.18688.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:31Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.71482s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:31\n",
      "INFO:tensorflow:Saving dict for global step 37000: average_loss = 100.07349, global_step = 37000, label/mean = 47.333332, loss = 100.07349, prediction/mean = 45.018898\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 37000: tf_wx_model\\model.ckpt-37000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:32Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.76036s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:33\n",
      "INFO:tensorflow:Saving dict for global step 37000: average_loss = 100.07349, global_step = 37000, label/mean = 47.333332, loss = 100.07349, prediction/mean = 45.018898\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 37000: tf_wx_model\\model.ckpt-37000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 37000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 51.134865, step = 37000\n",
      "INFO:tensorflow:global_step/sec: 134.663\n",
      "INFO:tensorflow:loss = 39.621624, step = 37100 (0.743 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 37200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 45.306026.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:37Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.78886s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:38\n",
      "INFO:tensorflow:Saving dict for global step 37200: average_loss = 100.10886, global_step = 37200, label/mean = 47.333332, loss = 100.10886, prediction/mean = 45.0379\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 37200: tf_wx_model\\model.ckpt-37200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:39Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.73757s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:40\n",
      "INFO:tensorflow:Saving dict for global step 37200: average_loss = 100.10886, global_step = 37200, label/mean = 47.333332, loss = 100.10886, prediction/mean = 45.0379\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 37200: tf_wx_model\\model.ckpt-37200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 37200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 37.720787, step = 37200\n",
      "INFO:tensorflow:global_step/sec: 124.323\n",
      "INFO:tensorflow:loss = 47.26901, step = 37300 (0.821 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 37400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 36.40655.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:45Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.75705s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:45\n",
      "INFO:tensorflow:Saving dict for global step 37400: average_loss = 100.77433, global_step = 37400, label/mean = 47.333332, loss = 100.77433, prediction/mean = 44.96425\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 37400: tf_wx_model\\model.ckpt-37400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:46Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.67578s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:47\n",
      "INFO:tensorflow:Saving dict for global step 37400: average_loss = 100.77433, global_step = 37400, label/mean = 47.333332, loss = 100.77433, prediction/mean = 44.96425\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 37400: tf_wx_model\\model.ckpt-37400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 37400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 56.292442, step = 37400\n",
      "INFO:tensorflow:global_step/sec: 136.718\n",
      "INFO:tensorflow:loss = 32.89186, step = 37500 (0.745 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 37600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 40.168133.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:51Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.77912s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:52\n",
      "INFO:tensorflow:Saving dict for global step 37600: average_loss = 100.974884, global_step = 37600, label/mean = 47.333332, loss = 100.974884, prediction/mean = 44.952896\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 37600: tf_wx_model\\model.ckpt-37600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:53Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.68148s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:54\n",
      "INFO:tensorflow:Saving dict for global step 37600: average_loss = 100.974884, global_step = 37600, label/mean = 47.333332, loss = 100.974884, prediction/mean = 44.952896\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 37600: tf_wx_model\\model.ckpt-37600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 37600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 27.483984, step = 37600\n",
      "INFO:tensorflow:global_step/sec: 127.088\n",
      "INFO:tensorflow:loss = 41.305035, step = 37700 (0.802 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 37800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 34.21824.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:58:59Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.66610s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:58:59\n",
      "INFO:tensorflow:Saving dict for global step 37800: average_loss = 100.58051, global_step = 37800, label/mean = 47.333332, loss = 100.58051, prediction/mean = 45.052624\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 37800: tf_wx_model\\model.ckpt-37800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:00Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.72743s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:01\n",
      "INFO:tensorflow:Saving dict for global step 37800: average_loss = 100.58051, global_step = 37800, label/mean = 47.333332, loss = 100.58051, prediction/mean = 45.052624\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 37800: tf_wx_model\\model.ckpt-37800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-37800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 37800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 26.846777, step = 37800\n",
      "INFO:tensorflow:global_step/sec: 130.916\n",
      "INFO:tensorflow:loss = 48.772263, step = 37900 (0.779 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 38000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 28.173162.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:06Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.91515s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:07\n",
      "INFO:tensorflow:Saving dict for global step 38000: average_loss = 100.9098, global_step = 38000, label/mean = 47.333332, loss = 100.9098, prediction/mean = 45.030712\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 38000: tf_wx_model\\model.ckpt-38000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:07Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.64487s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:08\n",
      "INFO:tensorflow:Saving dict for global step 38000: average_loss = 100.9098, global_step = 38000, label/mean = 47.333332, loss = 100.9098, prediction/mean = 45.030712\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 38000: tf_wx_model\\model.ckpt-38000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 38000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 21.306585, step = 38000\n",
      "INFO:tensorflow:global_step/sec: 124.54\n",
      "INFO:tensorflow:loss = 34.4676, step = 38100 (0.787 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 38200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 39.16099.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:13Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.57183s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:14\n",
      "INFO:tensorflow:Saving dict for global step 38200: average_loss = 100.52527, global_step = 38200, label/mean = 47.333332, loss = 100.52527, prediction/mean = 45.08738\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 38200: tf_wx_model\\model.ckpt-38200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:14Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.72666s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:15\n",
      "INFO:tensorflow:Saving dict for global step 38200: average_loss = 100.52527, global_step = 38200, label/mean = 47.333332, loss = 100.52527, prediction/mean = 45.08738\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 38200: tf_wx_model\\model.ckpt-38200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 38200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 48.360123, step = 38200\n",
      "INFO:tensorflow:global_step/sec: 126.412\n",
      "INFO:tensorflow:loss = 38.651417, step = 38300 (0.802 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 38400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 44.82585.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:20Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.73630s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:20\n",
      "INFO:tensorflow:Saving dict for global step 38400: average_loss = 100.504265, global_step = 38400, label/mean = 47.333332, loss = 100.504265, prediction/mean = 45.104744\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 38400: tf_wx_model\\model.ckpt-38400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:21Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.72245s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:22\n",
      "INFO:tensorflow:Saving dict for global step 38400: average_loss = 100.504265, global_step = 38400, label/mean = 47.333332, loss = 100.504265, prediction/mean = 45.104744\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 38400: tf_wx_model\\model.ckpt-38400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 38400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 38.184353, step = 38400\n",
      "INFO:tensorflow:global_step/sec: 127.492\n",
      "INFO:tensorflow:loss = 39.263634, step = 38500 (0.797 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 38600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 39.527462.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:27Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.66732s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:27\n",
      "INFO:tensorflow:Saving dict for global step 38600: average_loss = 100.632034, global_step = 38600, label/mean = 47.333332, loss = 100.632034, prediction/mean = 45.09606\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 38600: tf_wx_model\\model.ckpt-38600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:28Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.64047s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:29\n",
      "INFO:tensorflow:Saving dict for global step 38600: average_loss = 100.632034, global_step = 38600, label/mean = 47.333332, loss = 100.632034, prediction/mean = 45.09606\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 38600: tf_wx_model\\model.ckpt-38600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 38600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 38.587387, step = 38600\n",
      "INFO:tensorflow:global_step/sec: 131.067\n",
      "INFO:tensorflow:loss = 36.031857, step = 38700 (0.763 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 38800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 35.096107.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:34Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.96843s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:35\n",
      "INFO:tensorflow:Saving dict for global step 38800: average_loss = 101.348885, global_step = 38800, label/mean = 47.333332, loss = 101.348885, prediction/mean = 44.99439\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 38800: tf_wx_model\\model.ckpt-38800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:35Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.95425s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:36\n",
      "INFO:tensorflow:Saving dict for global step 38800: average_loss = 101.348885, global_step = 38800, label/mean = 47.333332, loss = 101.348885, prediction/mean = 44.99439\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 38800: tf_wx_model\\model.ckpt-38800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-38800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 38800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 26.643852, step = 38800\n",
      "INFO:tensorflow:global_step/sec: 107.744\n",
      "INFO:tensorflow:loss = 55.258022, step = 38900 (0.945 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 39000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 37.010185.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:41Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.42445s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:42\n",
      "INFO:tensorflow:Saving dict for global step 39000: average_loss = 100.83616, global_step = 39000, label/mean = 47.333332, loss = 100.83616, prediction/mean = 45.101807\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 39000: tf_wx_model\\model.ckpt-39000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:42Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.44987s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:42\n",
      "INFO:tensorflow:Saving dict for global step 39000: average_loss = 100.83616, global_step = 39000, label/mean = 47.333332, loss = 100.83616, prediction/mean = 45.101807\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 39000: tf_wx_model\\model.ckpt-39000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 39000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 40.00591, step = 39000\n",
      "INFO:tensorflow:global_step/sec: 133.742\n",
      "INFO:tensorflow:loss = 42.087223, step = 39100 (0.765 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 39200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 44.02529.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:46Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.72771s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:47\n",
      "INFO:tensorflow:Saving dict for global step 39200: average_loss = 101.013504, global_step = 39200, label/mean = 47.333332, loss = 101.013504, prediction/mean = 45.084538\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 39200: tf_wx_model\\model.ckpt-39200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:48Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.60090s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:48\n",
      "INFO:tensorflow:Saving dict for global step 39200: average_loss = 101.013504, global_step = 39200, label/mean = 47.333332, loss = 101.013504, prediction/mean = 45.084538\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 39200: tf_wx_model\\model.ckpt-39200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 39200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 43.88064, step = 39200\n",
      "INFO:tensorflow:global_step/sec: 134.388\n",
      "INFO:tensorflow:loss = 51.172314, step = 39300 (0.743 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 39400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 53.008533.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:53Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.58869s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:53\n",
      "INFO:tensorflow:Saving dict for global step 39400: average_loss = 101.42463, global_step = 39400, label/mean = 47.333332, loss = 101.42463, prediction/mean = 45.037647\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 39400: tf_wx_model\\model.ckpt-39400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:54Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.67401s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-17:59:55\n",
      "INFO:tensorflow:Saving dict for global step 39400: average_loss = 101.42463, global_step = 39400, label/mean = 47.333332, loss = 101.42463, prediction/mean = 45.037647\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 39400: tf_wx_model\\model.ckpt-39400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 39400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 34.77951, step = 39400\n",
      "INFO:tensorflow:global_step/sec: 139.743\n",
      "INFO:tensorflow:loss = 44.841778, step = 39500 (0.735 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 39600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 35.14745.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T17:59:59Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.65767s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:00\n",
      "INFO:tensorflow:Saving dict for global step 39600: average_loss = 101.50617, global_step = 39600, label/mean = 47.333332, loss = 101.50617, prediction/mean = 45.051083\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 39600: tf_wx_model\\model.ckpt-39600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:00Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.70136s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:01\n",
      "INFO:tensorflow:Saving dict for global step 39600: average_loss = 101.50617, global_step = 39600, label/mean = 47.333332, loss = 101.50617, prediction/mean = 45.051083\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 39600: tf_wx_model\\model.ckpt-39600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 39600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 33.50825, step = 39600\n",
      "INFO:tensorflow:global_step/sec: 141.897\n",
      "INFO:tensorflow:loss = 51.27853, step = 39700 (0.705 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 39800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 33.617374.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:05Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.62714s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:06\n",
      "INFO:tensorflow:Saving dict for global step 39800: average_loss = 102.027336, global_step = 39800, label/mean = 47.333332, loss = 102.027336, prediction/mean = 44.981983\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 39800: tf_wx_model\\model.ckpt-39800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:07Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.85288s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:08\n",
      "INFO:tensorflow:Saving dict for global step 39800: average_loss = 102.027336, global_step = 39800, label/mean = 47.333332, loss = 102.027336, prediction/mean = 44.981983\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 39800: tf_wx_model\\model.ckpt-39800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-39800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 39800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 38.62279, step = 39800\n",
      "INFO:tensorflow:global_step/sec: 161.089\n",
      "INFO:tensorflow:loss = 30.235205, step = 39900 (0.633 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 40000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 29.171354.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:11Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.61852s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:12\n",
      "INFO:tensorflow:Saving dict for global step 40000: average_loss = 102.28586, global_step = 40000, label/mean = 47.333332, loss = 102.28586, prediction/mean = 44.970047\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 40000: tf_wx_model\\model.ckpt-40000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:13Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.63490s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:13\n",
      "INFO:tensorflow:Saving dict for global step 40000: average_loss = 102.28586, global_step = 40000, label/mean = 47.333332, loss = 102.28586, prediction/mean = 44.970047\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 40000: tf_wx_model\\model.ckpt-40000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 40000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 45.31527, step = 40000\n",
      "INFO:tensorflow:global_step/sec: 143.805\n",
      "INFO:tensorflow:loss = 21.74222, step = 40100 (0.695 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 40200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 36.575947.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:18Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.67160s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:19\n",
      "INFO:tensorflow:Saving dict for global step 40200: average_loss = 102.2583, global_step = 40200, label/mean = 47.333332, loss = 102.2583, prediction/mean = 45.00516\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 40200: tf_wx_model\\model.ckpt-40200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:19Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 1.06568s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:20\n",
      "INFO:tensorflow:Saving dict for global step 40200: average_loss = 102.2583, global_step = 40200, label/mean = 47.333332, loss = 102.2583, prediction/mean = 45.00516\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 40200: tf_wx_model\\model.ckpt-40200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 40200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 32.4801, step = 40200\n",
      "INFO:tensorflow:global_step/sec: 116.956\n",
      "INFO:tensorflow:loss = 43.81201, step = 40300 (0.855 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 40400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 43.139565.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:25Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.91872s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:26\n",
      "INFO:tensorflow:Saving dict for global step 40400: average_loss = 102.14495, global_step = 40400, label/mean = 47.333332, loss = 102.14495, prediction/mean = 45.045895\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 40400: tf_wx_model\\model.ckpt-40400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:27Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.66171s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:27\n",
      "INFO:tensorflow:Saving dict for global step 40400: average_loss = 102.14495, global_step = 40400, label/mean = 47.333332, loss = 102.14495, prediction/mean = 45.045895\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 40400: tf_wx_model\\model.ckpt-40400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 40400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 50.336292, step = 40400\n",
      "INFO:tensorflow:global_step/sec: 121.581\n",
      "INFO:tensorflow:loss = 32.41685, step = 40500 (0.830 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 40600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 47.91147.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:32Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.60701s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:33\n",
      "INFO:tensorflow:Saving dict for global step 40600: average_loss = 102.08199, global_step = 40600, label/mean = 47.333332, loss = 102.08199, prediction/mean = 45.07781\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 40600: tf_wx_model\\model.ckpt-40600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:34Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.78861s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:34\n",
      "INFO:tensorflow:Saving dict for global step 40600: average_loss = 102.08199, global_step = 40600, label/mean = 47.333332, loss = 102.08199, prediction/mean = 45.07781\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 40600: tf_wx_model\\model.ckpt-40600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 40600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 36.401264, step = 40600\n",
      "INFO:tensorflow:global_step/sec: 135.225\n",
      "INFO:tensorflow:loss = 47.443966, step = 40700 (0.740 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 40800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 41.61299.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:40Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.86745s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:40\n",
      "INFO:tensorflow:Saving dict for global step 40800: average_loss = 102.60841, global_step = 40800, label/mean = 47.333332, loss = 102.60841, prediction/mean = 45.026306\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 40800: tf_wx_model\\model.ckpt-40800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:41Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.53123s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:42\n",
      "INFO:tensorflow:Saving dict for global step 40800: average_loss = 102.60841, global_step = 40800, label/mean = 47.333332, loss = 102.60841, prediction/mean = 45.026306\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 40800: tf_wx_model\\model.ckpt-40800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-40800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 40800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 41.205273, step = 40800\n",
      "INFO:tensorflow:global_step/sec: 126.886\n",
      "INFO:tensorflow:loss = 36.64589, step = 40900 (0.806 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 41000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 43.085087.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:46Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.77550s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:47\n",
      "INFO:tensorflow:Saving dict for global step 41000: average_loss = 103.46862, global_step = 41000, label/mean = 47.333332, loss = 103.46862, prediction/mean = 44.910465\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 41000: tf_wx_model\\model.ckpt-41000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:48Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.79573s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:49\n",
      "INFO:tensorflow:Saving dict for global step 41000: average_loss = 103.46862, global_step = 41000, label/mean = 47.333332, loss = 103.46862, prediction/mean = 44.910465\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 41000: tf_wx_model\\model.ckpt-41000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 41000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 27.02347, step = 41000\n",
      "INFO:tensorflow:global_step/sec: 111.414\n",
      "INFO:tensorflow:loss = 31.527334, step = 41100 (0.913 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 41200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 56.48859.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:53Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.68806s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:54\n",
      "INFO:tensorflow:Saving dict for global step 41200: average_loss = 102.419945, global_step = 41200, label/mean = 47.333332, loss = 102.419945, prediction/mean = 45.10274\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 41200: tf_wx_model\\model.ckpt-41200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:00:55Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.67663s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:00:56\n",
      "INFO:tensorflow:Saving dict for global step 41200: average_loss = 102.419945, global_step = 41200, label/mean = 47.333332, loss = 102.419945, prediction/mean = 45.10274\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 41200: tf_wx_model\\model.ckpt-41200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 41200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 41.065483, step = 41200\n",
      "INFO:tensorflow:global_step/sec: 126.015\n",
      "INFO:tensorflow:loss = 35.41406, step = 41300 (0.801 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 41400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 44.19348.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:01Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.69386s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:02\n",
      "INFO:tensorflow:Saving dict for global step 41400: average_loss = 103.288506, global_step = 41400, label/mean = 47.333332, loss = 103.288506, prediction/mean = 44.980503\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 41400: tf_wx_model\\model.ckpt-41400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:02Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.88684s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:03\n",
      "INFO:tensorflow:Saving dict for global step 41400: average_loss = 103.288506, global_step = 41400, label/mean = 47.333332, loss = 103.288506, prediction/mean = 44.980503\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 41400: tf_wx_model\\model.ckpt-41400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 41400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 39.970287, step = 41400\n",
      "INFO:tensorflow:global_step/sec: 136.385\n",
      "INFO:tensorflow:loss = 29.72708, step = 41500 (0.750 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 41600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 26.911602.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:07Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.82821s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:08\n",
      "INFO:tensorflow:Saving dict for global step 41600: average_loss = 102.99546, global_step = 41600, label/mean = 47.333332, loss = 102.99546, prediction/mean = 45.042774\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 41600: tf_wx_model\\model.ckpt-41600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:09Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.75460s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:10\n",
      "INFO:tensorflow:Saving dict for global step 41600: average_loss = 102.99546, global_step = 41600, label/mean = 47.333332, loss = 102.99546, prediction/mean = 45.042774\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 41600: tf_wx_model\\model.ckpt-41600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 41600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 49.821667, step = 41600\n",
      "INFO:tensorflow:global_step/sec: 121.281\n",
      "INFO:tensorflow:loss = 47.163376, step = 41700 (0.825 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 41800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 42.075542.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:14Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.58689s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:15\n",
      "INFO:tensorflow:Saving dict for global step 41800: average_loss = 103.19831, global_step = 41800, label/mean = 47.333332, loss = 103.19831, prediction/mean = 45.035255\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 41800: tf_wx_model\\model.ckpt-41800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:15Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.55138s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:16\n",
      "INFO:tensorflow:Saving dict for global step 41800: average_loss = 103.19831, global_step = 41800, label/mean = 47.333332, loss = 103.19831, prediction/mean = 45.035255\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 41800: tf_wx_model\\model.ckpt-41800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-41800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 41800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 38.87827, step = 41800\n",
      "INFO:tensorflow:global_step/sec: 141.886\n",
      "INFO:tensorflow:loss = 37.20407, step = 41900 (0.705 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 42000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 43.01045.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:20Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.89365s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:21\n",
      "INFO:tensorflow:Saving dict for global step 42000: average_loss = 103.15748, global_step = 42000, label/mean = 47.333332, loss = 103.15748, prediction/mean = 45.052353\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 42000: tf_wx_model\\model.ckpt-42000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:22Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.76364s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:23\n",
      "INFO:tensorflow:Saving dict for global step 42000: average_loss = 103.15748, global_step = 42000, label/mean = 47.333332, loss = 103.15748, prediction/mean = 45.052353\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 42000: tf_wx_model\\model.ckpt-42000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 42000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 32.978928, step = 42000\n",
      "INFO:tensorflow:global_step/sec: 96.821\n",
      "INFO:tensorflow:loss = 33.276295, step = 42100 (1.048 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 42200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 37.746674.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:28Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.58832s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:29\n",
      "INFO:tensorflow:Saving dict for global step 42200: average_loss = 103.68623, global_step = 42200, label/mean = 47.333332, loss = 103.68623, prediction/mean = 44.993855\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 42200: tf_wx_model\\model.ckpt-42200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:29Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.60010s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:30\n",
      "INFO:tensorflow:Saving dict for global step 42200: average_loss = 103.68623, global_step = 42200, label/mean = 47.333332, loss = 103.68623, prediction/mean = 44.993855\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 42200: tf_wx_model\\model.ckpt-42200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 42200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 38.75063, step = 42200\n",
      "INFO:tensorflow:global_step/sec: 129.917\n",
      "INFO:tensorflow:loss = 30.251421, step = 42300 (0.770 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 42400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 30.823479.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:34Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.68394s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:35\n",
      "INFO:tensorflow:Saving dict for global step 42400: average_loss = 103.65804, global_step = 42400, label/mean = 47.333332, loss = 103.65804, prediction/mean = 45.01379\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 42400: tf_wx_model\\model.ckpt-42400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:36Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 1.04086s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:37\n",
      "INFO:tensorflow:Saving dict for global step 42400: average_loss = 103.65804, global_step = 42400, label/mean = 47.333332, loss = 103.65804, prediction/mean = 45.01379\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 42400: tf_wx_model\\model.ckpt-42400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 42400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 55.400517, step = 42400\n",
      "INFO:tensorflow:global_step/sec: 118.283\n",
      "INFO:tensorflow:loss = 43.95677, step = 42500 (0.845 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 42600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 36.70836.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:42Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.75423s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:42\n",
      "INFO:tensorflow:Saving dict for global step 42600: average_loss = 103.48147, global_step = 42600, label/mean = 47.333332, loss = 103.48147, prediction/mean = 45.051125\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 42600: tf_wx_model\\model.ckpt-42600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:43Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.57679s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:44\n",
      "INFO:tensorflow:Saving dict for global step 42600: average_loss = 103.48147, global_step = 42600, label/mean = 47.333332, loss = 103.48147, prediction/mean = 45.051125\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 42600: tf_wx_model\\model.ckpt-42600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 42600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 47.17891, step = 42600\n",
      "INFO:tensorflow:global_step/sec: 81.2404\n",
      "INFO:tensorflow:loss = 59.848312, step = 42700 (1.248 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 42800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 48.75017.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:49Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.68515s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:50\n",
      "INFO:tensorflow:Saving dict for global step 42800: average_loss = 103.30443, global_step = 42800, label/mean = 47.333332, loss = 103.30443, prediction/mean = 45.107933\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 42800: tf_wx_model\\model.ckpt-42800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:51Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.76257s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:51\n",
      "INFO:tensorflow:Saving dict for global step 42800: average_loss = 103.30443, global_step = 42800, label/mean = 47.333332, loss = 103.30443, prediction/mean = 45.107933\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 42800: tf_wx_model\\model.ckpt-42800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-42800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 42800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 32.906075, step = 42800\n",
      "INFO:tensorflow:global_step/sec: 113.401\n",
      "INFO:tensorflow:loss = 31.154425, step = 42900 (0.902 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 43000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 32.52491.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:56Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.66117s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:57\n",
      "INFO:tensorflow:Saving dict for global step 43000: average_loss = 103.70073, global_step = 43000, label/mean = 47.333332, loss = 103.70073, prediction/mean = 45.05297\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 43000: tf_wx_model\\model.ckpt-43000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:01:57Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 1.02597s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:01:58\n",
      "INFO:tensorflow:Saving dict for global step 43000: average_loss = 103.70073, global_step = 43000, label/mean = 47.333332, loss = 103.70073, prediction/mean = 45.05297\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 43000: tf_wx_model\\model.ckpt-43000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 43000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 49.07821, step = 43000\n",
      "INFO:tensorflow:global_step/sec: 77.1682\n",
      "INFO:tensorflow:loss = 38.72722, step = 43100 (1.310 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 43200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 23.623919.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:04Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.61761s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:05\n",
      "INFO:tensorflow:Saving dict for global step 43200: average_loss = 104.42463, global_step = 43200, label/mean = 47.333332, loss = 104.42463, prediction/mean = 44.96298\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 43200: tf_wx_model\\model.ckpt-43200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:05Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.60455s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:06\n",
      "INFO:tensorflow:Saving dict for global step 43200: average_loss = 104.42463, global_step = 43200, label/mean = 47.333332, loss = 104.42463, prediction/mean = 44.96298\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 43200: tf_wx_model\\model.ckpt-43200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 43200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 33.367516, step = 43200\n",
      "INFO:tensorflow:global_step/sec: 114.085\n",
      "INFO:tensorflow:loss = 43.691757, step = 43300 (0.893 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 43400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 39.04874.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:12Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.82944s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:13\n",
      "INFO:tensorflow:Saving dict for global step 43400: average_loss = 104.05233, global_step = 43400, label/mean = 47.333332, loss = 104.05233, prediction/mean = 45.022343\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 43400: tf_wx_model\\model.ckpt-43400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:13Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.95013s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:14\n",
      "INFO:tensorflow:Saving dict for global step 43400: average_loss = 104.05233, global_step = 43400, label/mean = 47.333332, loss = 104.05233, prediction/mean = 45.022343\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 43400: tf_wx_model\\model.ckpt-43400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 43400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 47.060143, step = 43400\n",
      "INFO:tensorflow:global_step/sec: 118.727\n",
      "INFO:tensorflow:loss = 24.100769, step = 43500 (0.852 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 43600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 33.944527.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:20Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.91308s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:20\n",
      "INFO:tensorflow:Saving dict for global step 43600: average_loss = 104.15065, global_step = 43600, label/mean = 47.333332, loss = 104.15065, prediction/mean = 45.028183\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 43600: tf_wx_model\\model.ckpt-43600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:21Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.96994s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:22\n",
      "INFO:tensorflow:Saving dict for global step 43600: average_loss = 104.15065, global_step = 43600, label/mean = 47.333332, loss = 104.15065, prediction/mean = 45.028183\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 43600: tf_wx_model\\model.ckpt-43600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 43600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 32.848877, step = 43600\n",
      "INFO:tensorflow:global_step/sec: 63.6301\n",
      "INFO:tensorflow:loss = 43.55516, step = 43700 (1.577 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 43800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 24.782427.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:28Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.71331s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:29\n",
      "INFO:tensorflow:Saving dict for global step 43800: average_loss = 104.116295, global_step = 43800, label/mean = 47.333332, loss = 104.116295, prediction/mean = 45.036335\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 43800: tf_wx_model\\model.ckpt-43800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:30Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.68272s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:30\n",
      "INFO:tensorflow:Saving dict for global step 43800: average_loss = 104.116295, global_step = 43800, label/mean = 47.333332, loss = 104.116295, prediction/mean = 45.036335\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 43800: tf_wx_model\\model.ckpt-43800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-43800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 43800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 45.219467, step = 43800\n",
      "INFO:tensorflow:global_step/sec: 107.593\n",
      "INFO:tensorflow:loss = 23.64616, step = 43900 (0.939 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 44000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 30.46389.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:36Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.72170s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:36\n",
      "INFO:tensorflow:Saving dict for global step 44000: average_loss = 104.38267, global_step = 44000, label/mean = 47.333332, loss = 104.38267, prediction/mean = 45.008076\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 44000: tf_wx_model\\model.ckpt-44000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:37Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.71776s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:38\n",
      "INFO:tensorflow:Saving dict for global step 44000: average_loss = 104.38267, global_step = 44000, label/mean = 47.333332, loss = 104.38267, prediction/mean = 45.008076\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 44000: tf_wx_model\\model.ckpt-44000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 44000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 40.83093, step = 44000\n",
      "INFO:tensorflow:global_step/sec: 130.519\n",
      "INFO:tensorflow:loss = 43.669678, step = 44100 (0.775 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 44200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 37.36068.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:43Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.74445s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:44\n",
      "INFO:tensorflow:Saving dict for global step 44200: average_loss = 104.14572, global_step = 44200, label/mean = 47.333332, loss = 104.14572, prediction/mean = 45.04767\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 44200: tf_wx_model\\model.ckpt-44200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:44Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.90106s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:45\n",
      "INFO:tensorflow:Saving dict for global step 44200: average_loss = 104.14572, global_step = 44200, label/mean = 47.333332, loss = 104.14572, prediction/mean = 45.04767\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 44200: tf_wx_model\\model.ckpt-44200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 44200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 45.606148, step = 44200\n",
      "INFO:tensorflow:global_step/sec: 112.054\n",
      "INFO:tensorflow:loss = 34.78314, step = 44300 (0.877 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 44400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 36.08453.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:51Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.83308s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:51\n",
      "INFO:tensorflow:Saving dict for global step 44400: average_loss = 104.359215, global_step = 44400, label/mean = 47.333332, loss = 104.359215, prediction/mean = 45.049683\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 44400: tf_wx_model\\model.ckpt-44400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:52Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.63746s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:53\n",
      "INFO:tensorflow:Saving dict for global step 44400: average_loss = 104.359215, global_step = 44400, label/mean = 47.333332, loss = 104.359215, prediction/mean = 45.049683\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 44400: tf_wx_model\\model.ckpt-44400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 44400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 36.29089, step = 44400\n",
      "INFO:tensorflow:global_step/sec: 109.893\n",
      "INFO:tensorflow:loss = 23.290657, step = 44500 (0.912 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 44600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 40.82904.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:02:58Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.78868s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:02:59\n",
      "INFO:tensorflow:Saving dict for global step 44600: average_loss = 104.62629, global_step = 44600, label/mean = 47.333332, loss = 104.62629, prediction/mean = 45.024906\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 44600: tf_wx_model\\model.ckpt-44600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:00Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.80450s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:01\n",
      "INFO:tensorflow:Saving dict for global step 44600: average_loss = 104.62629, global_step = 44600, label/mean = 47.333332, loss = 104.62629, prediction/mean = 45.024906\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 44600: tf_wx_model\\model.ckpt-44600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 44600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 48.27248, step = 44600\n",
      "INFO:tensorflow:global_step/sec: 113.188\n",
      "INFO:tensorflow:loss = 31.269781, step = 44700 (0.891 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 44800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 40.198383.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:06Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.86400s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:07\n",
      "INFO:tensorflow:Saving dict for global step 44800: average_loss = 104.43351, global_step = 44800, label/mean = 47.333332, loss = 104.43351, prediction/mean = 45.069653\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 44800: tf_wx_model\\model.ckpt-44800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:07Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.64099s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:08\n",
      "INFO:tensorflow:Saving dict for global step 44800: average_loss = 104.43351, global_step = 44800, label/mean = 47.333332, loss = 104.43351, prediction/mean = 45.069653\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 44800: tf_wx_model\\model.ckpt-44800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-44800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 44800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 29.79612, step = 44800\n",
      "INFO:tensorflow:global_step/sec: 111.779\n",
      "INFO:tensorflow:loss = 36.0468, step = 44900 (0.910 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 45000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 36.731884.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:13Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.79277s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:14\n",
      "INFO:tensorflow:Saving dict for global step 45000: average_loss = 104.97561, global_step = 45000, label/mean = 47.333332, loss = 104.97561, prediction/mean = 44.998405\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 45000: tf_wx_model\\model.ckpt-45000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:15Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.57955s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:15\n",
      "INFO:tensorflow:Saving dict for global step 45000: average_loss = 104.97561, global_step = 45000, label/mean = 47.333332, loss = 104.97561, prediction/mean = 44.998405\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 45000: tf_wx_model\\model.ckpt-45000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 45000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 30.758066, step = 45000\n",
      "INFO:tensorflow:global_step/sec: 146.018\n",
      "INFO:tensorflow:loss = 32.543095, step = 45100 (0.685 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 45200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 31.692688.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:20Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.85172s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:21\n",
      "INFO:tensorflow:Saving dict for global step 45200: average_loss = 105.35975, global_step = 45200, label/mean = 47.333332, loss = 105.35975, prediction/mean = 44.970257\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 45200: tf_wx_model\\model.ckpt-45200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:22Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.70179s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:22\n",
      "INFO:tensorflow:Saving dict for global step 45200: average_loss = 105.35975, global_step = 45200, label/mean = 47.333332, loss = 105.35975, prediction/mean = 44.970257\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 45200: tf_wx_model\\model.ckpt-45200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 45200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 32.18633, step = 45200\n",
      "INFO:tensorflow:global_step/sec: 124.944\n",
      "INFO:tensorflow:loss = 24.81869, step = 45300 (0.816 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 45400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 50.12561.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:27Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.66373s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:27\n",
      "INFO:tensorflow:Saving dict for global step 45400: average_loss = 105.09124, global_step = 45400, label/mean = 47.333332, loss = 105.09124, prediction/mean = 45.03935\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 45400: tf_wx_model\\model.ckpt-45400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:28Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.60438s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:29\n",
      "INFO:tensorflow:Saving dict for global step 45400: average_loss = 105.09124, global_step = 45400, label/mean = 47.333332, loss = 105.09124, prediction/mean = 45.03935\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 45400: tf_wx_model\\model.ckpt-45400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 45400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 41.124535, step = 45400\n",
      "INFO:tensorflow:global_step/sec: 107.328\n",
      "INFO:tensorflow:loss = 40.933086, step = 45500 (0.936 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 45600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 32.81343.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:34Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.82167s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:35\n",
      "INFO:tensorflow:Saving dict for global step 45600: average_loss = 105.668, global_step = 45600, label/mean = 47.333332, loss = 105.668, prediction/mean = 44.937695\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 45600: tf_wx_model\\model.ckpt-45600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:36Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.92075s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:37\n",
      "INFO:tensorflow:Saving dict for global step 45600: average_loss = 105.668, global_step = 45600, label/mean = 47.333332, loss = 105.668, prediction/mean = 44.937695\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 45600: tf_wx_model\\model.ckpt-45600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 45600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 39.592316, step = 45600\n",
      "INFO:tensorflow:global_step/sec: 116.272\n",
      "INFO:tensorflow:loss = 47.83887, step = 45700 (0.873 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 45800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 58.254837.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:41Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.55779s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:42\n",
      "INFO:tensorflow:Saving dict for global step 45800: average_loss = 105.12075, global_step = 45800, label/mean = 47.333332, loss = 105.12075, prediction/mean = 45.037548\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 45800: tf_wx_model\\model.ckpt-45800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:43Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.62395s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:43\n",
      "INFO:tensorflow:Saving dict for global step 45800: average_loss = 105.12075, global_step = 45800, label/mean = 47.333332, loss = 105.12075, prediction/mean = 45.037548\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 45800: tf_wx_model\\model.ckpt-45800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-45800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 45800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 43.51689, step = 45800\n",
      "INFO:tensorflow:global_step/sec: 120.683\n",
      "INFO:tensorflow:loss = 41.467525, step = 45900 (0.829 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 46000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 37.83862.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:48Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.71413s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:49\n",
      "INFO:tensorflow:Saving dict for global step 46000: average_loss = 105.33821, global_step = 46000, label/mean = 47.333332, loss = 105.33821, prediction/mean = 45.017464\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 46000: tf_wx_model\\model.ckpt-46000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:50Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.60472s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:50\n",
      "INFO:tensorflow:Saving dict for global step 46000: average_loss = 105.33821, global_step = 46000, label/mean = 47.333332, loss = 105.33821, prediction/mean = 45.017464\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 46000: tf_wx_model\\model.ckpt-46000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 46000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 41.979935, step = 46000\n",
      "INFO:tensorflow:global_step/sec: 129.621\n",
      "INFO:tensorflow:loss = 31.162422, step = 46100 (0.775 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 46200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 36.569416.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:55Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.87960s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:56\n",
      "INFO:tensorflow:Saving dict for global step 46200: average_loss = 106.022934, global_step = 46200, label/mean = 47.333332, loss = 106.022934, prediction/mean = 44.92311\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 46200: tf_wx_model\\model.ckpt-46200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:03:56Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.68135s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:03:57\n",
      "INFO:tensorflow:Saving dict for global step 46200: average_loss = 106.022934, global_step = 46200, label/mean = 47.333332, loss = 106.022934, prediction/mean = 44.92311\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 46200: tf_wx_model\\model.ckpt-46200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 46200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 48.148182, step = 46200\n",
      "INFO:tensorflow:global_step/sec: 152.982\n",
      "INFO:tensorflow:loss = 32.492043, step = 46300 (0.663 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 46400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 48.814312.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:02Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.75763s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:02\n",
      "INFO:tensorflow:Saving dict for global step 46400: average_loss = 105.677505, global_step = 46400, label/mean = 47.333332, loss = 105.677505, prediction/mean = 45.003807\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 46400: tf_wx_model\\model.ckpt-46400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:03Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.90239s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:04\n",
      "INFO:tensorflow:Saving dict for global step 46400: average_loss = 105.677505, global_step = 46400, label/mean = 47.333332, loss = 105.677505, prediction/mean = 45.003807\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 46400: tf_wx_model\\model.ckpt-46400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 46400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 31.758278, step = 46400\n",
      "INFO:tensorflow:global_step/sec: 133.625\n",
      "INFO:tensorflow:loss = 38.844643, step = 46500 (0.747 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 46600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 35.22885.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:08Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.67993s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:09\n",
      "INFO:tensorflow:Saving dict for global step 46600: average_loss = 105.772835, global_step = 46600, label/mean = 47.333332, loss = 105.772835, prediction/mean = 44.99246\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 46600: tf_wx_model\\model.ckpt-46600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:10Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.71894s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:10\n",
      "INFO:tensorflow:Saving dict for global step 46600: average_loss = 105.772835, global_step = 46600, label/mean = 47.333332, loss = 105.772835, prediction/mean = 44.99246\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 46600: tf_wx_model\\model.ckpt-46600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 46600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 29.19817, step = 46600\n",
      "INFO:tensorflow:global_step/sec: 114.915\n",
      "INFO:tensorflow:loss = 40.87904, step = 46700 (0.869 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 46800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 27.378576.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:15Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.76453s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:16\n",
      "INFO:tensorflow:Saving dict for global step 46800: average_loss = 105.88265, global_step = 46800, label/mean = 47.333332, loss = 105.88265, prediction/mean = 44.98915\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 46800: tf_wx_model\\model.ckpt-46800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:16Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.72754s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:17\n",
      "INFO:tensorflow:Saving dict for global step 46800: average_loss = 105.88265, global_step = 46800, label/mean = 47.333332, loss = 105.88265, prediction/mean = 44.98915\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 46800: tf_wx_model\\model.ckpt-46800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-46800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 46800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 34.538326, step = 46800\n",
      "INFO:tensorflow:global_step/sec: 103.008\n",
      "INFO:tensorflow:loss = 24.88867, step = 46900 (0.971 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 47000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 29.800571.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:22Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 1.06824s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:23\n",
      "INFO:tensorflow:Saving dict for global step 47000: average_loss = 105.97546, global_step = 47000, label/mean = 47.333332, loss = 105.97546, prediction/mean = 44.99821\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 47000: tf_wx_model\\model.ckpt-47000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:24Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.67756s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:25\n",
      "INFO:tensorflow:Saving dict for global step 47000: average_loss = 105.97546, global_step = 47000, label/mean = 47.333332, loss = 105.97546, prediction/mean = 44.99821\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 47000: tf_wx_model\\model.ckpt-47000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 47000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 37.97368, step = 47000\n",
      "INFO:tensorflow:global_step/sec: 120.549\n",
      "INFO:tensorflow:loss = 28.70058, step = 47100 (0.838 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 47200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 33.238888.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:30Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.68399s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:30\n",
      "INFO:tensorflow:Saving dict for global step 47200: average_loss = 105.89153, global_step = 47200, label/mean = 47.333332, loss = 105.89153, prediction/mean = 45.022495\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 47200: tf_wx_model\\model.ckpt-47200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:31Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.50900s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:31\n",
      "INFO:tensorflow:Saving dict for global step 47200: average_loss = 105.89153, global_step = 47200, label/mean = 47.333332, loss = 105.89153, prediction/mean = 45.022495\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 47200: tf_wx_model\\model.ckpt-47200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 47200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 24.037844, step = 47200\n",
      "INFO:tensorflow:global_step/sec: 121.531\n",
      "INFO:tensorflow:loss = 42.070633, step = 47300 (0.820 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 47400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 47.415905.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:37Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.78067s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:37\n",
      "INFO:tensorflow:Saving dict for global step 47400: average_loss = 106.024574, global_step = 47400, label/mean = 47.333332, loss = 106.024574, prediction/mean = 45.006516\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 47400: tf_wx_model\\model.ckpt-47400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:38Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.74119s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:39\n",
      "INFO:tensorflow:Saving dict for global step 47400: average_loss = 106.024574, global_step = 47400, label/mean = 47.333332, loss = 106.024574, prediction/mean = 45.006516\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 47400: tf_wx_model\\model.ckpt-47400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 47400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 34.20337, step = 47400\n",
      "INFO:tensorflow:global_step/sec: 147.339\n",
      "INFO:tensorflow:loss = 43.528164, step = 47500 (0.679 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 47600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 39.945183.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:44Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.72340s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:45\n",
      "INFO:tensorflow:Saving dict for global step 47600: average_loss = 105.88778, global_step = 47600, label/mean = 47.333332, loss = 105.88778, prediction/mean = 45.04081\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 47600: tf_wx_model\\model.ckpt-47600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:45Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.77977s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:46\n",
      "INFO:tensorflow:Saving dict for global step 47600: average_loss = 105.88778, global_step = 47600, label/mean = 47.333332, loss = 105.88778, prediction/mean = 45.04081\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 47600: tf_wx_model\\model.ckpt-47600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 47600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 32.21396, step = 47600\n",
      "INFO:tensorflow:global_step/sec: 115.449\n",
      "INFO:tensorflow:loss = 33.680134, step = 47700 (0.875 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 47800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 35.75439.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:51Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.68044s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:51\n",
      "INFO:tensorflow:Saving dict for global step 47800: average_loss = 106.442955, global_step = 47800, label/mean = 47.333332, loss = 106.442955, prediction/mean = 44.95677\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 47800: tf_wx_model\\model.ckpt-47800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:52Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.67368s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:53\n",
      "INFO:tensorflow:Saving dict for global step 47800: average_loss = 106.442955, global_step = 47800, label/mean = 47.333332, loss = 106.442955, prediction/mean = 44.95677\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 47800: tf_wx_model\\model.ckpt-47800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-47800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 47800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 28.38633, step = 47800\n",
      "INFO:tensorflow:global_step/sec: 123.419\n",
      "INFO:tensorflow:loss = 39.70686, step = 47900 (0.810 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 48000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 29.166618.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:58Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.71319s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:04:58\n",
      "INFO:tensorflow:Saving dict for global step 48000: average_loss = 106.58044, global_step = 48000, label/mean = 47.333332, loss = 106.58044, prediction/mean = 44.959072\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 48000: tf_wx_model\\model.ckpt-48000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:04:59Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.78489s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:00\n",
      "INFO:tensorflow:Saving dict for global step 48000: average_loss = 106.58044, global_step = 48000, label/mean = 47.333332, loss = 106.58044, prediction/mean = 44.959072\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 48000: tf_wx_model\\model.ckpt-48000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 48000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 33.75592, step = 48000\n",
      "INFO:tensorflow:global_step/sec: 144.923\n",
      "INFO:tensorflow:loss = 47.21067, step = 48100 (0.699 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 48200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 23.271788.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:04Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.71285s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:05\n",
      "INFO:tensorflow:Saving dict for global step 48200: average_loss = 106.09987, global_step = 48200, label/mean = 47.333332, loss = 106.09987, prediction/mean = 45.065323\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 48200: tf_wx_model\\model.ckpt-48200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:06Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.76665s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:06\n",
      "INFO:tensorflow:Saving dict for global step 48200: average_loss = 106.09987, global_step = 48200, label/mean = 47.333332, loss = 106.09987, prediction/mean = 45.065323\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 48200: tf_wx_model\\model.ckpt-48200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 48200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 39.166374, step = 48200\n",
      "INFO:tensorflow:global_step/sec: 141.556\n",
      "INFO:tensorflow:loss = 28.790672, step = 48300 (0.713 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 48400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 32.685818.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:12Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.86119s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:13\n",
      "INFO:tensorflow:Saving dict for global step 48400: average_loss = 106.51782, global_step = 48400, label/mean = 47.333332, loss = 106.51782, prediction/mean = 44.99957\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 48400: tf_wx_model\\model.ckpt-48400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:13Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.75047s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:14\n",
      "INFO:tensorflow:Saving dict for global step 48400: average_loss = 106.51782, global_step = 48400, label/mean = 47.333332, loss = 106.51782, prediction/mean = 44.99957\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 48400: tf_wx_model\\model.ckpt-48400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 48400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 40.806053, step = 48400\n",
      "INFO:tensorflow:global_step/sec: 135.205\n",
      "INFO:tensorflow:loss = 47.742123, step = 48500 (0.741 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 48600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 44.896805.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:19Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.59543s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:20\n",
      "INFO:tensorflow:Saving dict for global step 48600: average_loss = 106.48771, global_step = 48600, label/mean = 47.333332, loss = 106.48771, prediction/mean = 45.013176\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 48600: tf_wx_model\\model.ckpt-48600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:20Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.86594s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:21\n",
      "INFO:tensorflow:Saving dict for global step 48600: average_loss = 106.48771, global_step = 48600, label/mean = 47.333332, loss = 106.48771, prediction/mean = 45.013176\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 48600: tf_wx_model\\model.ckpt-48600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 48600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 24.85002, step = 48600\n",
      "INFO:tensorflow:global_step/sec: 121.843\n",
      "INFO:tensorflow:loss = 21.477522, step = 48700 (0.821 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 48800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 22.933561.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:26Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.82518s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:27\n",
      "INFO:tensorflow:Saving dict for global step 48800: average_loss = 106.71299, global_step = 48800, label/mean = 47.333332, loss = 106.71299, prediction/mean = 45.00379\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 48800: tf_wx_model\\model.ckpt-48800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:28Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.63577s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:28\n",
      "INFO:tensorflow:Saving dict for global step 48800: average_loss = 106.71299, global_step = 48800, label/mean = 47.333332, loss = 106.71299, prediction/mean = 45.00379\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 48800: tf_wx_model\\model.ckpt-48800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-48800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 48800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 42.65188, step = 48800\n",
      "INFO:tensorflow:global_step/sec: 150.338\n",
      "INFO:tensorflow:loss = 47.918, step = 48900 (0.665 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 49000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 21.623707.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:33Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.54766s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:34\n",
      "INFO:tensorflow:Saving dict for global step 49000: average_loss = 106.82402, global_step = 49000, label/mean = 47.333332, loss = 106.82402, prediction/mean = 45.00392\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 49000: tf_wx_model\\model.ckpt-49000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:34Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.80605s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:35\n",
      "INFO:tensorflow:Saving dict for global step 49000: average_loss = 106.82402, global_step = 49000, label/mean = 47.333332, loss = 106.82402, prediction/mean = 45.00392\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 49000: tf_wx_model\\model.ckpt-49000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 49000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 41.50292, step = 49000\n",
      "INFO:tensorflow:global_step/sec: 98.0971\n",
      "INFO:tensorflow:loss = 26.940193, step = 49100 (1.036 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 49200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 33.334152.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:40Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.66708s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:41\n",
      "INFO:tensorflow:Saving dict for global step 49200: average_loss = 106.755775, global_step = 49200, label/mean = 47.333332, loss = 106.755775, prediction/mean = 45.041607\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 49200: tf_wx_model\\model.ckpt-49200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:42Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.61210s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:42\n",
      "INFO:tensorflow:Saving dict for global step 49200: average_loss = 106.755775, global_step = 49200, label/mean = 47.333332, loss = 106.755775, prediction/mean = 45.041607\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 49200: tf_wx_model\\model.ckpt-49200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 49200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 36.72039, step = 49200\n",
      "INFO:tensorflow:global_step/sec: 125.496\n",
      "INFO:tensorflow:loss = 39.564034, step = 49300 (0.800 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 49400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 36.3996.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:47Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.69325s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:48\n",
      "INFO:tensorflow:Saving dict for global step 49400: average_loss = 107.28732, global_step = 49400, label/mean = 47.333332, loss = 107.28732, prediction/mean = 44.962498\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 49400: tf_wx_model\\model.ckpt-49400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:48Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.60456s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:49\n",
      "INFO:tensorflow:Saving dict for global step 49400: average_loss = 107.28732, global_step = 49400, label/mean = 47.333332, loss = 107.28732, prediction/mean = 44.962498\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 49400: tf_wx_model\\model.ckpt-49400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 49400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 40.26848, step = 49400\n",
      "INFO:tensorflow:global_step/sec: 123.541\n",
      "INFO:tensorflow:loss = 38.064278, step = 49500 (0.826 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 49600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 28.324268.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:53Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.73979s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:54\n",
      "INFO:tensorflow:Saving dict for global step 49600: average_loss = 107.19603, global_step = 49600, label/mean = 47.333332, loss = 107.19603, prediction/mean = 44.99906\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 49600: tf_wx_model\\model.ckpt-49600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:05:55Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.57611s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:05:56\n",
      "INFO:tensorflow:Saving dict for global step 49600: average_loss = 107.19603, global_step = 49600, label/mean = 47.333332, loss = 107.19603, prediction/mean = 44.99906\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 49600: tf_wx_model\\model.ckpt-49600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 49600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 20.196196, step = 49600\n",
      "INFO:tensorflow:global_step/sec: 130.588\n",
      "INFO:tensorflow:loss = 40.595432, step = 49700 (0.766 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 49800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 35.58895.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:00Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.77425s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:01\n",
      "INFO:tensorflow:Saving dict for global step 49800: average_loss = 107.07643, global_step = 49800, label/mean = 47.333332, loss = 107.07643, prediction/mean = 45.034737\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 49800: tf_wx_model\\model.ckpt-49800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:02Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.66340s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:02\n",
      "INFO:tensorflow:Saving dict for global step 49800: average_loss = 107.07643, global_step = 49800, label/mean = 47.333332, loss = 107.07643, prediction/mean = 45.034737\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 49800: tf_wx_model\\model.ckpt-49800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-49800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 49800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 26.773062, step = 49800\n",
      "INFO:tensorflow:global_step/sec: 116.814\n",
      "INFO:tensorflow:loss = 39.292786, step = 49900 (0.869 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 50000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 34.653603.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:07Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.79806s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:08\n",
      "INFO:tensorflow:Saving dict for global step 50000: average_loss = 107.530075, global_step = 50000, label/mean = 47.333332, loss = 107.530075, prediction/mean = 44.958096\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50000: tf_wx_model\\model.ckpt-50000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:08Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.71130s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:09\n",
      "INFO:tensorflow:Saving dict for global step 50000: average_loss = 107.530075, global_step = 50000, label/mean = 47.333332, loss = 107.530075, prediction/mean = 44.958096\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50000: tf_wx_model\\model.ckpt-50000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 50000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 47.423363, step = 50000\n",
      "INFO:tensorflow:global_step/sec: 143.585\n",
      "INFO:tensorflow:loss = 33.136627, step = 50100 (0.696 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 50200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 38.68557.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:14Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.70464s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:15\n",
      "INFO:tensorflow:Saving dict for global step 50200: average_loss = 107.18635, global_step = 50200, label/mean = 47.333332, loss = 107.18635, prediction/mean = 45.05123\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50200: tf_wx_model\\model.ckpt-50200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:15Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.86385s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:16\n",
      "INFO:tensorflow:Saving dict for global step 50200: average_loss = 107.18635, global_step = 50200, label/mean = 47.333332, loss = 107.18635, prediction/mean = 45.05123\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50200: tf_wx_model\\model.ckpt-50200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 50200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 23.713972, step = 50200\n",
      "INFO:tensorflow:global_step/sec: 137.51\n",
      "INFO:tensorflow:loss = 24.088459, step = 50300 (0.751 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 50400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 33.322083.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:21Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.62006s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:22\n",
      "INFO:tensorflow:Saving dict for global step 50400: average_loss = 107.42117, global_step = 50400, label/mean = 47.333332, loss = 107.42117, prediction/mean = 45.020634\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50400: tf_wx_model\\model.ckpt-50400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:22Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.60311s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:23\n",
      "INFO:tensorflow:Saving dict for global step 50400: average_loss = 107.42117, global_step = 50400, label/mean = 47.333332, loss = 107.42117, prediction/mean = 45.020634\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50400: tf_wx_model\\model.ckpt-50400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 50400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 39.140835, step = 50400\n",
      "INFO:tensorflow:global_step/sec: 107.219\n",
      "INFO:tensorflow:loss = 29.479614, step = 50500 (0.933 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 50600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 29.034042.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:28Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.68770s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:29\n",
      "INFO:tensorflow:Saving dict for global step 50600: average_loss = 107.176, global_step = 50600, label/mean = 47.333332, loss = 107.176, prediction/mean = 45.068604\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50600: tf_wx_model\\model.ckpt-50600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:29Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.83245s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:30\n",
      "INFO:tensorflow:Saving dict for global step 50600: average_loss = 107.176, global_step = 50600, label/mean = 47.333332, loss = 107.176, prediction/mean = 45.068604\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50600: tf_wx_model\\model.ckpt-50600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 50600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 33.949, step = 50600\n",
      "INFO:tensorflow:global_step/sec: 103.541\n",
      "INFO:tensorflow:loss = 38.889084, step = 50700 (0.966 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 50800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 32.652916.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:35Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.67975s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:36\n",
      "INFO:tensorflow:Saving dict for global step 50800: average_loss = 107.947525, global_step = 50800, label/mean = 47.333332, loss = 107.947525, prediction/mean = 44.976547\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50800: tf_wx_model\\model.ckpt-50800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:37Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.87030s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:38\n",
      "INFO:tensorflow:Saving dict for global step 50800: average_loss = 107.947525, global_step = 50800, label/mean = 47.333332, loss = 107.947525, prediction/mean = 44.976547\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 50800: tf_wx_model\\model.ckpt-50800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-50800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 50800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 26.145998, step = 50800\n",
      "INFO:tensorflow:global_step/sec: 156.136\n",
      "INFO:tensorflow:loss = 46.542103, step = 50900 (0.640 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 51000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 28.172466.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:42Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.72338s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:43\n",
      "INFO:tensorflow:Saving dict for global step 51000: average_loss = 107.80605, global_step = 51000, label/mean = 47.333332, loss = 107.80605, prediction/mean = 45.019478\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 51000: tf_wx_model\\model.ckpt-51000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:43Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.68261s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:44\n",
      "INFO:tensorflow:Saving dict for global step 51000: average_loss = 107.80605, global_step = 51000, label/mean = 47.333332, loss = 107.80605, prediction/mean = 45.019478\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 51000: tf_wx_model\\model.ckpt-51000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 51000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 26.316729, step = 51000\n",
      "INFO:tensorflow:global_step/sec: 116.314\n",
      "INFO:tensorflow:loss = 26.202888, step = 51100 (0.860 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 51200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 48.42756.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:49Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.81718s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:50\n",
      "INFO:tensorflow:Saving dict for global step 51200: average_loss = 107.61162, global_step = 51200, label/mean = 47.333332, loss = 107.61162, prediction/mean = 45.056007\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 51200: tf_wx_model\\model.ckpt-51200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:50Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.70994s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:51\n",
      "INFO:tensorflow:Saving dict for global step 51200: average_loss = 107.61162, global_step = 51200, label/mean = 47.333332, loss = 107.61162, prediction/mean = 45.056007\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 51200: tf_wx_model\\model.ckpt-51200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 51200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 46.384483, step = 51200\n",
      "INFO:tensorflow:global_step/sec: 106.907\n",
      "INFO:tensorflow:loss = 24.845264, step = 51300 (0.935 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 51400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 26.05058.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:56Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.63310s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:57\n",
      "INFO:tensorflow:Saving dict for global step 51400: average_loss = 107.6952, global_step = 51400, label/mean = 47.333332, loss = 107.6952, prediction/mean = 45.061447\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 51400: tf_wx_model\\model.ckpt-51400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:06:58Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.65003s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:06:58\n",
      "INFO:tensorflow:Saving dict for global step 51400: average_loss = 107.6952, global_step = 51400, label/mean = 47.333332, loss = 107.6952, prediction/mean = 45.061447\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 51400: tf_wx_model\\model.ckpt-51400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 51400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 33.263924, step = 51400\n",
      "INFO:tensorflow:global_step/sec: 134.807\n",
      "INFO:tensorflow:loss = 28.30787, step = 51500 (0.757 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 51600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 30.785833.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:03Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.82840s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:04\n",
      "INFO:tensorflow:Saving dict for global step 51600: average_loss = 107.70505, global_step = 51600, label/mean = 47.333332, loss = 107.70505, prediction/mean = 45.069954\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 51600: tf_wx_model\\model.ckpt-51600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:04Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.58495s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:05\n",
      "INFO:tensorflow:Saving dict for global step 51600: average_loss = 107.70505, global_step = 51600, label/mean = 47.333332, loss = 107.70505, prediction/mean = 45.069954\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 51600: tf_wx_model\\model.ckpt-51600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 51600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 33.00215, step = 51600\n",
      "INFO:tensorflow:global_step/sec: 126.499\n",
      "INFO:tensorflow:loss = 27.377012, step = 51700 (0.806 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 51800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 31.058104.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:10Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.67733s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:10\n",
      "INFO:tensorflow:Saving dict for global step 51800: average_loss = 108.2897, global_step = 51800, label/mean = 47.333332, loss = 108.2897, prediction/mean = 44.997234\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 51800: tf_wx_model\\model.ckpt-51800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:11Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.65375s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:12\n",
      "INFO:tensorflow:Saving dict for global step 51800: average_loss = 108.2897, global_step = 51800, label/mean = 47.333332, loss = 108.2897, prediction/mean = 44.997234\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 51800: tf_wx_model\\model.ckpt-51800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-51800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 51800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 22.436255, step = 51800\n",
      "INFO:tensorflow:global_step/sec: 132.597\n",
      "INFO:tensorflow:loss = 47.81529, step = 51900 (0.754 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 52000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 38.222412.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:17Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.74912s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:18\n",
      "INFO:tensorflow:Saving dict for global step 52000: average_loss = 107.73742, global_step = 52000, label/mean = 47.333332, loss = 107.73742, prediction/mean = 45.086082\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 52000: tf_wx_model\\model.ckpt-52000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:18Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.76415s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:19\n",
      "INFO:tensorflow:Saving dict for global step 52000: average_loss = 107.73742, global_step = 52000, label/mean = 47.333332, loss = 107.73742, prediction/mean = 45.086082\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 52000: tf_wx_model\\model.ckpt-52000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 52000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 47.307953, step = 52000\n",
      "INFO:tensorflow:global_step/sec: 114.915\n",
      "INFO:tensorflow:loss = 31.232233, step = 52100 (0.897 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 52200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 37.970894.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:24Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.81461s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:25\n",
      "INFO:tensorflow:Saving dict for global step 52200: average_loss = 108.16017, global_step = 52200, label/mean = 47.333332, loss = 108.16017, prediction/mean = 45.041897\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 52200: tf_wx_model\\model.ckpt-52200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:26Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.84635s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:27\n",
      "INFO:tensorflow:Saving dict for global step 52200: average_loss = 108.16017, global_step = 52200, label/mean = 47.333332, loss = 108.16017, prediction/mean = 45.041897\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 52200: tf_wx_model\\model.ckpt-52200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 52200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 31.372326, step = 52200\n",
      "INFO:tensorflow:global_step/sec: 103.044\n",
      "INFO:tensorflow:loss = 29.877785, step = 52300 (0.973 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 52400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 37.09229.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:34Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.73300s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:35\n",
      "INFO:tensorflow:Saving dict for global step 52400: average_loss = 108.07085, global_step = 52400, label/mean = 47.333332, loss = 108.07085, prediction/mean = 45.064365\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 52400: tf_wx_model\\model.ckpt-52400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:35Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.74664s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:36\n",
      "INFO:tensorflow:Saving dict for global step 52400: average_loss = 108.07085, global_step = 52400, label/mean = 47.333332, loss = 108.07085, prediction/mean = 45.064365\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 52400: tf_wx_model\\model.ckpt-52400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 52400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 31.05664, step = 52400\n",
      "INFO:tensorflow:global_step/sec: 80.3705\n",
      "INFO:tensorflow:loss = 32.399517, step = 52500 (1.260 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 52600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 35.693913.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:43Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.70582s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:43\n",
      "INFO:tensorflow:Saving dict for global step 52600: average_loss = 108.35495, global_step = 52600, label/mean = 47.333332, loss = 108.35495, prediction/mean = 45.027378\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 52600: tf_wx_model\\model.ckpt-52600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:44Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.58400s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:45\n",
      "INFO:tensorflow:Saving dict for global step 52600: average_loss = 108.35495, global_step = 52600, label/mean = 47.333332, loss = 108.35495, prediction/mean = 45.027378\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 52600: tf_wx_model\\model.ckpt-52600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 52600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 42.787586, step = 52600\n",
      "INFO:tensorflow:global_step/sec: 117.037\n",
      "INFO:tensorflow:loss = 29.077564, step = 52700 (0.862 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 52800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 26.782848.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:49Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.80110s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:50\n",
      "INFO:tensorflow:Saving dict for global step 52800: average_loss = 108.417694, global_step = 52800, label/mean = 47.333332, loss = 108.417694, prediction/mean = 45.02919\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 52800: tf_wx_model\\model.ckpt-52800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:51Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.61401s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:51\n",
      "INFO:tensorflow:Saving dict for global step 52800: average_loss = 108.417694, global_step = 52800, label/mean = 47.333332, loss = 108.417694, prediction/mean = 45.02919\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 52800: tf_wx_model\\model.ckpt-52800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-52800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 52800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 42.906216, step = 52800\n",
      "INFO:tensorflow:global_step/sec: 104.36\n",
      "INFO:tensorflow:loss = 36.227905, step = 52900 (0.968 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 53000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 42.433765.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:57Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.71631s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:58\n",
      "INFO:tensorflow:Saving dict for global step 53000: average_loss = 108.51099, global_step = 53000, label/mean = 47.333332, loss = 108.51099, prediction/mean = 45.042164\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 53000: tf_wx_model\\model.ckpt-53000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:07:58Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.78494s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:07:59\n",
      "INFO:tensorflow:Saving dict for global step 53000: average_loss = 108.51099, global_step = 53000, label/mean = 47.333332, loss = 108.51099, prediction/mean = 45.042164\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 53000: tf_wx_model\\model.ckpt-53000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 53000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 19.58526, step = 53000\n",
      "INFO:tensorflow:global_step/sec: 115.049\n",
      "INFO:tensorflow:loss = 29.061396, step = 53100 (0.886 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 53200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 33.791695.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:04Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.75909s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:05\n",
      "INFO:tensorflow:Saving dict for global step 53200: average_loss = 108.10234, global_step = 53200, label/mean = 47.333332, loss = 108.10234, prediction/mean = 45.10048\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 53200: tf_wx_model\\model.ckpt-53200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:06Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.89196s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:07\n",
      "INFO:tensorflow:Saving dict for global step 53200: average_loss = 108.10234, global_step = 53200, label/mean = 47.333332, loss = 108.10234, prediction/mean = 45.10048\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 53200: tf_wx_model\\model.ckpt-53200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 53200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 30.710945, step = 53200\n",
      "INFO:tensorflow:global_step/sec: 92.8891\n",
      "INFO:tensorflow:loss = 41.259125, step = 53300 (1.093 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 53400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 33.130936.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:13Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.62565s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:13\n",
      "INFO:tensorflow:Saving dict for global step 53400: average_loss = 108.728645, global_step = 53400, label/mean = 47.333332, loss = 108.728645, prediction/mean = 45.01683\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 53400: tf_wx_model\\model.ckpt-53400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:14Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.70424s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:15\n",
      "INFO:tensorflow:Saving dict for global step 53400: average_loss = 108.728645, global_step = 53400, label/mean = 47.333332, loss = 108.728645, prediction/mean = 45.01683\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 53400: tf_wx_model\\model.ckpt-53400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 53400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 36.632915, step = 53400\n",
      "INFO:tensorflow:global_step/sec: 100.666\n",
      "INFO:tensorflow:loss = 48.331738, step = 53500 (0.993 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 53600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 19.672436.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:21Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.80105s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:21\n",
      "INFO:tensorflow:Saving dict for global step 53600: average_loss = 108.51564, global_step = 53600, label/mean = 47.333332, loss = 108.51564, prediction/mean = 45.05244\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 53600: tf_wx_model\\model.ckpt-53600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:22Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.55706s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:23\n",
      "INFO:tensorflow:Saving dict for global step 53600: average_loss = 108.51564, global_step = 53600, label/mean = 47.333332, loss = 108.51564, prediction/mean = 45.05244\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 53600: tf_wx_model\\model.ckpt-53600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 53600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 31.043657, step = 53600\n",
      "INFO:tensorflow:global_step/sec: 89.8974\n",
      "INFO:tensorflow:loss = 30.605911, step = 53700 (1.117 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 53800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 28.17457.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:28Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.75456s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:29\n",
      "INFO:tensorflow:Saving dict for global step 53800: average_loss = 108.5354, global_step = 53800, label/mean = 47.333332, loss = 108.5354, prediction/mean = 45.062393\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 53800: tf_wx_model\\model.ckpt-53800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:30Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.64629s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:31\n",
      "INFO:tensorflow:Saving dict for global step 53800: average_loss = 108.5354, global_step = 53800, label/mean = 47.333332, loss = 108.5354, prediction/mean = 45.062393\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 53800: tf_wx_model\\model.ckpt-53800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-53800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 53800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 23.107246, step = 53800\n",
      "INFO:tensorflow:global_step/sec: 92.4216\n",
      "INFO:tensorflow:loss = 24.915203, step = 53900 (1.096 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 54000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 41.97515.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:36Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.74995s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:37\n",
      "INFO:tensorflow:Saving dict for global step 54000: average_loss = 109.10327, global_step = 54000, label/mean = 47.333332, loss = 109.10327, prediction/mean = 44.998055\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 54000: tf_wx_model\\model.ckpt-54000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:38Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.78423s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:39\n",
      "INFO:tensorflow:Saving dict for global step 54000: average_loss = 109.10327, global_step = 54000, label/mean = 47.333332, loss = 109.10327, prediction/mean = 44.998055\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 54000: tf_wx_model\\model.ckpt-54000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 54000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 47.156998, step = 54000\n",
      "INFO:tensorflow:global_step/sec: 105.031\n",
      "INFO:tensorflow:loss = 27.162373, step = 54100 (0.953 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 54200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 35.86281.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:44Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.68617s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:45\n",
      "INFO:tensorflow:Saving dict for global step 54200: average_loss = 108.93029, global_step = 54200, label/mean = 47.333332, loss = 108.93029, prediction/mean = 45.03632\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 54200: tf_wx_model\\model.ckpt-54200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:45Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.61982s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:46\n",
      "INFO:tensorflow:Saving dict for global step 54200: average_loss = 108.93029, global_step = 54200, label/mean = 47.333332, loss = 108.93029, prediction/mean = 45.03632\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 54200: tf_wx_model\\model.ckpt-54200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 54200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 35.312256, step = 54200\n",
      "INFO:tensorflow:global_step/sec: 89.3564\n",
      "INFO:tensorflow:loss = 35.890476, step = 54300 (1.136 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 54400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 34.758175.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:52Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.63410s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:53\n",
      "INFO:tensorflow:Saving dict for global step 54400: average_loss = 109.156685, global_step = 54400, label/mean = 47.333332, loss = 109.156685, prediction/mean = 45.015293\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 54400: tf_wx_model\\model.ckpt-54400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:08:53Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.63106s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:08:54\n",
      "INFO:tensorflow:Saving dict for global step 54400: average_loss = 109.156685, global_step = 54400, label/mean = 47.333332, loss = 109.156685, prediction/mean = 45.015293\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 54400: tf_wx_model\\model.ckpt-54400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 54400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 37.07518, step = 54400\n",
      "INFO:tensorflow:global_step/sec: 71.4032\n",
      "INFO:tensorflow:loss = 25.734966, step = 54500 (1.400 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 54600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 25.84027.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:01Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.71370s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:01\n",
      "INFO:tensorflow:Saving dict for global step 54600: average_loss = 109.140205, global_step = 54600, label/mean = 47.333332, loss = 109.140205, prediction/mean = 45.034832\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 54600: tf_wx_model\\model.ckpt-54600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:02Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.57519s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:03\n",
      "INFO:tensorflow:Saving dict for global step 54600: average_loss = 109.140205, global_step = 54600, label/mean = 47.333332, loss = 109.140205, prediction/mean = 45.034832\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 54600: tf_wx_model\\model.ckpt-54600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 54600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 21.898405, step = 54600\n",
      "INFO:tensorflow:global_step/sec: 118.985\n",
      "INFO:tensorflow:loss = 21.355156, step = 54700 (0.856 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 54800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 34.70126.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:08Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.76333s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:09\n",
      "INFO:tensorflow:Saving dict for global step 54800: average_loss = 109.35757, global_step = 54800, label/mean = 47.333332, loss = 109.35757, prediction/mean = 45.01618\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 54800: tf_wx_model\\model.ckpt-54800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:10Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.59916s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:10\n",
      "INFO:tensorflow:Saving dict for global step 54800: average_loss = 109.35757, global_step = 54800, label/mean = 47.333332, loss = 109.35757, prediction/mean = 45.01618\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 54800: tf_wx_model\\model.ckpt-54800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-54800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 54800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 28.144365, step = 54800\n",
      "INFO:tensorflow:global_step/sec: 88.5199\n",
      "INFO:tensorflow:loss = 36.027992, step = 54900 (1.130 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 55000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 30.167324.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:16Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.60988s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:17\n",
      "INFO:tensorflow:Saving dict for global step 55000: average_loss = 109.522675, global_step = 55000, label/mean = 47.333332, loss = 109.522675, prediction/mean = 44.985714\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 55000: tf_wx_model\\model.ckpt-55000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:17Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.62950s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:18\n",
      "INFO:tensorflow:Saving dict for global step 55000: average_loss = 109.522675, global_step = 55000, label/mean = 47.333332, loss = 109.522675, prediction/mean = 44.985714\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 55000: tf_wx_model\\model.ckpt-55000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 55000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 24.056107, step = 55000\n",
      "INFO:tensorflow:global_step/sec: 88.4394\n",
      "INFO:tensorflow:loss = 45.8083, step = 55100 (1.138 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 55200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 31.190363.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:24Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 1.06902s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:25\n",
      "INFO:tensorflow:Saving dict for global step 55200: average_loss = 109.43538, global_step = 55200, label/mean = 47.333332, loss = 109.43538, prediction/mean = 45.007847\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 55200: tf_wx_model\\model.ckpt-55200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:26Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.64216s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:26\n",
      "INFO:tensorflow:Saving dict for global step 55200: average_loss = 109.43538, global_step = 55200, label/mean = 47.333332, loss = 109.43538, prediction/mean = 45.007847\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 55200: tf_wx_model\\model.ckpt-55200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 55200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 35.29408, step = 55200\n",
      "INFO:tensorflow:global_step/sec: 115.519\n",
      "INFO:tensorflow:loss = 33.489517, step = 55300 (0.866 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 55400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 24.432877.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:31Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.76295s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:32\n",
      "INFO:tensorflow:Saving dict for global step 55400: average_loss = 109.69489, global_step = 55400, label/mean = 47.333332, loss = 109.69489, prediction/mean = 44.96279\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 55400: tf_wx_model\\model.ckpt-55400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:32Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.67030s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:33\n",
      "INFO:tensorflow:Saving dict for global step 55400: average_loss = 109.69489, global_step = 55400, label/mean = 47.333332, loss = 109.69489, prediction/mean = 44.96279\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 55400: tf_wx_model\\model.ckpt-55400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 55400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 24.559303, step = 55400\n",
      "INFO:tensorflow:global_step/sec: 116.248\n",
      "INFO:tensorflow:loss = 46.81336, step = 55500 (0.877 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 55600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 40.982758.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:38Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.60490s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:39\n",
      "INFO:tensorflow:Saving dict for global step 55600: average_loss = 109.415985, global_step = 55600, label/mean = 47.333332, loss = 109.415985, prediction/mean = 45.04494\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 55600: tf_wx_model\\model.ckpt-55600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:39Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.70184s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:40\n",
      "INFO:tensorflow:Saving dict for global step 55600: average_loss = 109.415985, global_step = 55600, label/mean = 47.333332, loss = 109.415985, prediction/mean = 45.04494\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 55600: tf_wx_model\\model.ckpt-55600\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55600\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 55600 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 46.153866, step = 55600\n",
      "INFO:tensorflow:global_step/sec: 112.32\n",
      "INFO:tensorflow:loss = 28.344124, step = 55700 (0.890 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 55800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 37.144863.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:46Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.80725s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:47\n",
      "INFO:tensorflow:Saving dict for global step 55800: average_loss = 109.69138, global_step = 55800, label/mean = 47.333332, loss = 109.69138, prediction/mean = 44.999355\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 55800: tf_wx_model\\model.ckpt-55800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:47Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.60632s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:48\n",
      "INFO:tensorflow:Saving dict for global step 55800: average_loss = 109.69138, global_step = 55800, label/mean = 47.333332, loss = 109.69138, prediction/mean = 44.999355\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 55800: tf_wx_model\\model.ckpt-55800\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-55800\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 55800 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 22.651375, step = 55800\n",
      "INFO:tensorflow:global_step/sec: 97.9065\n",
      "INFO:tensorflow:loss = 39.012314, step = 55900 (1.027 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 56000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 34.674973.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:54Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-56000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.69603s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:55\n",
      "INFO:tensorflow:Saving dict for global step 56000: average_loss = 109.569046, global_step = 56000, label/mean = 47.333332, loss = 109.569046, prediction/mean = 45.02876\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 56000: tf_wx_model\\model.ckpt-56000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:09:55Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-56000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.67361s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:09:56\n",
      "INFO:tensorflow:Saving dict for global step 56000: average_loss = 109.569046, global_step = 56000, label/mean = 47.333332, loss = 109.569046, prediction/mean = 45.02876\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 56000: tf_wx_model\\model.ckpt-56000\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-56000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 56000 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 22.449762, step = 56000\n",
      "INFO:tensorflow:global_step/sec: 78.7136\n",
      "INFO:tensorflow:loss = 23.92264, step = 56100 (1.276 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 56200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 23.518972.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:10:02Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-56200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.57447s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:10:03\n",
      "INFO:tensorflow:Saving dict for global step 56200: average_loss = 109.3472, global_step = 56200, label/mean = 47.333332, loss = 109.3472, prediction/mean = 45.06281\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 56200: tf_wx_model\\model.ckpt-56200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:10:03Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-56200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.65169s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:10:04\n",
      "INFO:tensorflow:Saving dict for global step 56200: average_loss = 109.3472, global_step = 56200, label/mean = 47.333332, loss = 109.3472, prediction/mean = 45.06281\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 56200: tf_wx_model\\model.ckpt-56200\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-56200\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 56200 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:loss = 30.025408, step = 56200\n",
      "INFO:tensorflow:global_step/sec: 94.1855\n",
      "INFO:tensorflow:loss = 37.95257, step = 56300 (1.062 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 56400 into tf_wx_model\\model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 17.755423.\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:10:09Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-56400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Inference Time : 0.73389s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:10:10\n",
      "INFO:tensorflow:Saving dict for global step 56400: average_loss = 109.468506, global_step = 56400, label/mean = 47.333332, loss = 109.468506, prediction/mean = 45.053402\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 56400: tf_wx_model\\model.ckpt-56400\n",
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2020-06-14T18:10:10Z\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-56400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Inference Time : 0.62377s\n",
      "INFO:tensorflow:Finished evaluation at 2020-06-14-18:10:11\n",
      "INFO:tensorflow:Saving dict for global step 56400: average_loss = 109.468506, global_step = 56400, label/mean = 47.333332, loss = 109.468506, prediction/mean = 45.053402\n",
      "INFO:tensorflow:Saving 'checkpoint_path' summary for global step 56400: tf_wx_model\\model.ckpt-56400\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'average_loss': 99.79377,\n",
       " 'label/mean': 47.333332,\n",
       " 'loss': 99.79377,\n",
       " 'prediction/mean': 45.038815,\n",
       " 'global_step': 36600}"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def wx_input_fn(X, y=None, num_epochs=None, shuffle=True, batch_size=50):\n",
    "    return tf.compat.v1.estimator.inputs.pandas_input_fn(x=X,\n",
    "                                               y=y,\n",
    "                                               num_epochs=num_epochs,\n",
    "                                               shuffle=shuffle,\n",
    "                                               batch_size=batch_size)\n",
    "\n",
    "\n",
    "evaluations = []\n",
    "STEPS = 200\n",
    "for i in range(100):\n",
    "    regressor.train(input_fn=wx_input_fn(X_train, y=y_train), steps=STEPS)\n",
    "    evaluation = regressor.evaluate(input_fn=wx_input_fn(X_val, y_val,\n",
    "                                                         num_epochs=1,\n",
    "                                                         shuffle=False),\n",
    "                                    steps=1)\n",
    "    evaluations.append(regressor.evaluate(input_fn=wx_input_fn(X_val,\n",
    "                                                               y_val,\n",
    "                                                               num_epochs=1,\n",
    "\n",
    "                                                               shuffle=False)))\n",
    "\n",
    "evaluations[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA0cAAAJQCAYAAABFHh8UAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzs3X903eldH/j3px4T7kKKEjJJxwrDTEpQD9QFty7QurTZ5ASFlBLXlEKWLSmEzXah3W57VsU+bIFuy7GpS+HQlLKB5tdCA5S4SpYh64ZMQ2gggKdK4oStyUBoGDnNJARB6Wpbx3n2D13ZspGuZEnf+/P1OkdH9z73e+999PiOxm8/z/N5qrUWAACAWfcHRt0BAACAcSAcAQAARDgCAABIIhwBAAAkEY4AAACSCEcAAABJhCMAAIAkwhEAAEAS4QgAACBJct+oO3AQz3jGM9pDDz006m4AAABj6rHHHvtYa+3+vVw70eHooYceypUrV0bdDQAAYExV1X/Y67WW1QEAAEQ4AgAASCIcAQAAJBGOAAAAknQYjqrq1VX1ZFW9b0vbV1fV+6vqk1V18q7rz1XV41V1raoWu+oXAADAdrqcOXptkhfd1fa+JGeSvGNrY1V9XpKvTfL5/ef8QFUd6bBvAAAAd+gsHLXW3pHk43e1/T+ttWvbXP6SJD/WWvsvrbUPJnk8yRd11TcAAIC7jcueo/kkv7nl/hP9NgAAgKEYl3BU27S1bS+sekVVXamqKx/96Ec77hYAADArxiUcPZHks7bcf3aS69td2Fp7VWvtZGvt5P333z+UzgEAANNvXMLRm5N8bVU9paoeTvLcJL804j4BAAAz5L6uXriq3pDkeUmeUVVPJPmObBRo+CdJ7k/ySFW9u7W22Fp7f1X9RJJfSfKJJN/SWrvZVd8AAADu1lk4aq29dIeH/tUO139Xku/qqj8AAACDjMuyOgAAgJESjgAAACIcAQAAJBGOAAAAkghHAAAASYQjAACAJMIRAABAkg7POQIAALq1vLKai5ev5fraeo7N9bK0uJDTJ+Y7e960E44AAGACLa+s5tylq1m/cTNJsrq2nnOXribJwKCz3+fNAsvqAABgAl28fO1WwNm0fuNmLl6+1snzZoFwBAAAE+j62vo9tR/0ebNAOAIAgAl0bK53T+0Hfd4sEI4AAGACLS0upHf0yB1tvaNHsrS40MnzZoGCDAAAMIE2iyfca9W5vTxvVqvZVWtt1H3Yt5MnT7YrV66MuhsAADA17q5ml2zMLJ0/c3wiA1JVPdZaO7mXay2rAwAAbpnlanaW1QEAwIiN0zK2Wa5mZ+YIAABGaHMZ2+raelpuH8q6vLI6kv7McjU74QgAAEZo3JaxHaSa3fLKak5deDQPn30kpy48OrKAt1+W1QEAwAiN2zK2/VbBu7uQw+YM2NbXHHfCEQAAjNCxuV5WtwlCm8vYRrEf6fSJ+Xt+j0EzYJMSjiyrAwCAERq0jO0g+5GGvcRt3GbA9kM4AgCAETp9Yj7nzxzP/FwvlWR+rnfrTKH97kcaRZGHaSjkYFkdAACM2E7L2PY7GzOKJW5LiwvbHh67l0IO40I4AgCAIdjP3qHd9iPtZBRL3PZbyGGcCEcAANCx/VZy2+9szH5D1V4MCnn7KeQwTuw5AgCAju1379Cg/UiDHOSsokHG7cDaw2bmCAAAOnaQZW77mY3paonbNJTrHkQ4AgCAjnW5zG0nXSxxm4Zy3YNYVgcAAB3rapnbsE1Due5BhCMAAOjYfvcOjZtpCXk7sawOAACGYNIruSXTUa57EOEIAADYs2kIeTuxrA4AACBmjgAAYM8GHYDK5BOOAABgDzYPQN0852fzANRkY6mZ4DT5hCMAANiDQQegJhkYnJgMwhEAADNnP7M8gw5AHRSchKPJoSADAAAzZXN53Oraelpuz/Isr6wOfN6gA1AHBScmh3AEAMBM2W153E4GHYA6KDgxOSyrAwBgpux3lme3A1C37jlKbgcnJodwBADATDk218vqNkFoL7M8Ox2AultwYjIIRwAAzJSlxYVOZnl2Ck5MDuEIAICZYpaHnQhHAADMHLM8bEc4AgCALfZzBhLTQTgCAIC+zTOQNvcjbZ6BlERAmgHOOQIAgL79noHEdBCOAACgb79nIDEdhCMAAOjb6ayjvZyBxOQTjgAAoG9pcSG9o0fuaDuMM5CYDAoyAABAnzOQZptwBAAAWzgDaXZZVgcAABDhCAAAIIlldQAAjLHllVX7fxga4QgAgLG0vLKac5eu3jqUdXVtPecuXU0SAYlOWFYHAMBYunj52q1gtGn9xs1cvHxtRD1i2glHAACMpetr6/fUDgclHAEAMJaOzfXuqR0OSjgCAGAsLS0upHf0yB1tvaNHsrS4MKIeMe0UZAAAYCxtFl1QrY5hEY4AABhbp0/MC0MMjWV1AAAAEY4AAACSWFYHAMCUWl5ZtV+JeyIcAQDQuS6CyqDXXF5ZzblLV28dIru6tp5zl64miYDEjiyrAwCgU5tBZXVtPS23g8ryympnr3nx8rVbwWjT+o2buXj52gF+EqadcAQAQKe6CCq7veb1tfVtn7dTOyTCEQAAHesiqOz2msfmets+vlM7JMIRAAAd6yKo7PaaS4sL6R09csdjvaNHsrS4sO/3ZPoJRwAAHIrlldWcuvBoHj77SE5dePTW/p8ugspur3n6xHzOnzme+bleKsn8XC/nzxxXjIGBVKsDAODA9lId7jCr1e3lNU+fmBeGuCfVWht1H/bt5MmT7cqVK6PuBgDAzDt14dGsbrMPaH6ul3eeff4IegQbquqx1trJvVxr5ggAgD3b6Wwh1eGYBsIRAAB7Mmjp3LG53rYzR6rDMUkUZAAAYE8GnS2kOhzTwMwRAMCE2mmJW1cGLZ3rougCDJtwBAAwgfZSHe6w7bZ0TnU4Jp1ldQAAE2jQEreuWDrHtDNzBAAwgUZRHc7SOaadcAQAMIFGVR3O0jmmmWV1AAATyBI3OHxmjgAAJpAlbnD4hCMAgAm13yVuwy4BDpNCOAIAmCGjKAEOk8KeIwCAGTKKEuAwKcwcAQDMkFGUAN+NZX6MC+EIAGCG7FYCfNhBxTI/xolldQAAM2RQCfDNoLK6tp6W20FleWW1s/5Y5sc4EY4AAGbI6RPzOX/meObneqkk83O9nD9zPKdPzI8kqIzjMj9mV2fL6qrq1Um+IsmTrbU/2m97epIfT/JQkt9I8pdba79dVZ+R5EeSPNjv0z9qrb2mq74BAMyynUqAjyKo7LbMD4apy5mj1yZ50V1tZ5O8rbX23CRv699Pkm9J8iuttS9I8rwk31NVn9Jh3wAAuMtOgaTLoDJomR8MW2fhqLX2jiQfv6v5JUle17/9uiSnNy9P8tSqqiSf3n/eJ7rqGwAAv98ogsqgZX4wbMOuVves1tqHk6S19uGqema//ZVJ3pzkepKnJvma1tonh9w3AICBpr3k9ObPMuyfcadlfjBs41LKezHJu5M8P8kfTvLWqvq51trv3n1hVb0iySuS5MEHHxxqJwGA2TUrJacFFWbZsKvVfaSqHkiS/vcn++3fkORS2/B4kg8m+SPbvUBr7VWttZOttZP333//UDoNAKDkNEy/YYejNyd5Wf/2y5K8qX/7Q0lekCRV9awkC0l+fch9AwDYkZLTMP06C0dV9YYkv5BkoaqeqKqXJ7mQ5IVV9YEkL+zfT5K/n+RPV9XVbFSx+9bW2se66hsAwL0aRSU3YLg623PUWnvpDg+9YJtrryf5sq76AgBwUEuLC3fsOUqUnIZpMy4FGQAAxtqoKrkBwyMcAQDs0aBKbtNe5htmgXAEAHBAs1LmG6bdsKvVAQBMHWW+YToIRwAAB6TMN0wHy+oAAA7o2Fwvq9sEoVGW+bYHCu6dmSMAgANaWlxI7+iRO9oOq8z38spqTl14NA+ffSSnLjya5ZXVPT3n3KWrWV1bT8vtPVB7eS7MMjNHAAAHdNAy3zvN8uy30MOgPVBmj2BnwhEAwCEYVOZ7kEEBaL8hxx4o2B/L6gAARmhQANpvyNlpr9Mo90DBJBCOAABGaFAA2m/I6XIPFEwz4QgAoGODiioMCkD7DTmnT8zn/JnjmZ/rpZLMz/Vy/sxx+41gF/YcAQB0aLeiCkuLC3c8ntwOQAcp9LDfPVAwy4QjAIAO7VZUYbcAJOTA8AhHAAAd2ktRBQEIxoNwBABMpJ3OBho3x+Z6Wd0mIKkcB+NHQQYAYOJs7uNZXVtPy+19PFsLHYwLleNgcghHAMDEGbSPZ9yoHAeTw7I6AGDi7Pdw1FGxpwgmg5kjAGDi7PdwVIBBhCMAYOLYxwN0wbI6AGDiHORw1GRyKt0BwyUcAQATab/7eDYr3W0WdNisdLf5msDssqwOAJgpk1TpDhgu4QgAmCmTVukOGB7hCACYKSrdATsRjgCAmaLSHbATBRkAgM6NU3W4g1a6A6aXcAQAdGocq8Ptt9IdMN0sqwMAOqU6HDAphCMAoFOqwwGTQjgCADqlOhwwKew5AgA6tbS4cMeeo+TO6nBdFWsYpyIQwGQQjgCATg2qDtdVsYaDvK5QBbOrWmuj7sO+nTx5sl25cmXU3QAA9unUhUezus3eo/m5Xt559vlDf927Q1WyMct1/sxxAQkmVFU91lo7uZdr7TkCAEamq2IN+31dlfVgtglHAMDIdFWsYb+vq7IezDbhCAAYmaXFhfSOHrmjbWuxhmG/rsp6MNsUZAAARmZQsYZk/8URdnvdnexWWQ+YbgoyAABjaVTFEVSrg+lyLwUZzBwBAGNpUHGELsPK6RPzwhDMKHuOAICxpDgCMGzCEQAwlhRHAIZNOAIAxlJXlewAdmLPEQAwlvZbcQ5gv4QjAGBsKY4ADJNldQAAABGOAAAAkghHAAAASYQjAACAJAoyAAB7tLyyqnIcMNWEIwBgV8srqzl36WrWb9xMkqyurefcpatJIiABU8OyOgBgVxcvX7sVjDat37iZi5evjahHAIdPOAIAdnV9bf2e2gEmkWV1AMCujs31srpNEDo217t1254kYNKZOQIAdrW0uJDe0SN3tPWOHsnS4kKS23uSVtfW03J7T9LyyuoIeguwP2aOAIBbdpr92ZwB2mlmaNCeJLNHwKQQjgCAJLtXpNsaku5mTxIwDSyrAwCSHKwi3da9R3tpBxhHwhEAkORgsz+77UkCmATCEQCQ5GCzP6dPzOf8meOZn+ulkszP9XL+zHH7jYCJYs8RAJBkY/Zn656j5N5mfwbtSQKYBMIRAIyxYZ4dtFtFOoBpJxwBwJjarXpcF8z+ALPMniMAGFMHqR4HwL0TjgBgTDk7CGC4LKsDgDF1bK6X1W2C0F6qxw1zrxLAtDBzBABjar9nB23uVVpdW0/L7b1Kyyurtx4/deHRPHz2kZy68OitdoBZZ+YIAMbUfqvH7bZXadhFHgAmRbXWRt2HfTt58mS7cuXKqLsBAPvWxfK3h88+ku3+717Zeane/Fwv7zz7/AO9L8A4qqrHWmsn93KtZXUAMCK7LX/br532JB2b6ynyADCAcAQAI9JVqe5Be5UGBSeAWWfPEQCMSFezOLvtVdq65yjZW5EHgFkgHAHAiBykVPduTp+Y33bv0n6LPADMAuEIAEZkaXFhJLM4OwUngFknHAHAiJjFARgvwhEAjJBZHIDxoVodAABAhCMAAIAkwhEAAEAS4QgAACCJcAQAAJBEOAIAAEgiHAEAACQRjgAAAJIIRwAAAEmS+0bdAQBgf5ZXVnPx8rVcX1vPsblelhYXcvrE/Ki7BTCxhCMAmEDLK6s5d+lq1m/cTJKsrq3n3KWrSSIgAeyTZXUAMIEuXr52KxhtWr9xMxcvXxtRjwAmn3AEABPo+tr6PbUDsDvL6gAYW5O0p2bYfT0218vqNkHo2Fyvs/cEmHadzRxV1aur6smqet+WtqdX1Vur6gP970/b8tjzqurdVfX+qvrZrvoFwGTY3FOzuraeltt7apZXVkfdtd9nFH1dWlxI7+iRO9p6R49kaXGhs/cEmHZdLqt7bZIX3dV2NsnbWmvPTfK2/v1U1VySH0jyla21z0/y1R32C4AJMEl7akbR19Mn5nP+zPHMz/VSSebnejl/5vjYzqwBTILOltW11t5RVQ/d1fySJM/r335dkrcn+dYk/12SS621D/Wf+2RX/QJgMoxqT81+lseNqq+nT8wLQwCHaNgFGZ7VWvtwkvS/P7Pf/rlJnlZVb6+qx6rq63d6gap6RVVdqaorH/3oR4fQZQBGYae9M13uqdnv8rhR9BWAwzcu1eruS/Inkvz5JItJ/m5Vfe52F7bWXtVaO9laO3n//fcPs48ADNEo9tTsd3mc/T8A02HY1eo+UlUPtNY+XFUPJNlcPvdEko+11v5zkv9cVe9I8gVJfnXI/QNgTGwuFxtmBbj9Lo8bRV8BOHzDDkdvTvKyJBf639/Ub39TkldW1X1JPiXJFyf53iH3DYAx09Wemp32Fe1WHnvQfqRBfZ2kkuQAs6xaa928cNUbslF84RlJPpLkO5IsJ/mJJA8m+VCSr26tfbx//VKSb0jyySQ/3Fr7vt3e4+TJk+3KlSud9B+A6bS5r2jr8rne0SM5f+Z4kuzrsUFBZ9D7CUgA3auqx1prJ/d0bVfhaBiEIwDu1akLj247OzQ/18s7zz5/x1me3Z633/cDoFv3Eo6GvawOAEZqt31FOy2P2+9+pFGV+Qbg3o1LtToAGIr9lt0e9vMAGD7hCICZst+y28N+HgDDZ1kdAFNnt6pyyb2X3R728wAYPgUZAJgqqsMBsNW9FGSwrA6AqXLx8rU7glGSrN+4mYuXr42oRwBMCuEIgKmiOhwA+2XPEQATaad9RcfmetueK6Q6HAC7EY4AOBSDiiB08V5b9xWtrq3n3KWrSTaqw22350h1OAB2Y1kdAAe2GVZW19bTcjusLK+sdvJ+g/YVnT4xn/Nnjmd+rpdKMj/XU4wBgD0xcwTAge0WVg7bbvuKTp+YF4YAuGdmjgA4sGEXQdhp/5B9RQAchHAEwIENO6wsLS6kd/TIHW32FQFwUMIRAAc27LBiXxEAXbDnCIAD2wwlw6pWt/mewhAAh0k4AuBQCCsATDrL6gAAACIcAQAAJBGOAAAAkghHAAAASYQjAACAJMIRAABAEuEIAAAgiXOOANhieWV1qAe5AsA4EY4ASLIRjM5dupr1GzeTJKtr6zl36WqSdBqQBDIAxoVldQAkSS5evnYrGG1av3EzFy9f6+w9NwPZ6tp6Wm4HsuWV1c7eEwB2IhwBkCS5vrZ+T+2HYRSBDAB2IhwBkCQ5Nte7p/bDMIpABgA7EY4ASJIsLS6kd/TIHW29o0eytLjQ2XuOIpABwE6EIwCSbBRdOH/meObneqkk83O9nD9zvNPiCKMIZACwE9XqALjl9In5oVaK23wv1eoAGAfCEQAjNexABgA7sawOAAAge5g5qqpnJ/naJF+a5FiS9STvS/JIkre01j7ZaQ8BAACGYGA4qqrXJJlP8lNJvjvJk0k+NcnnJnlRkm+rqrOttXd03VEAAIAu7TZz9D2ttfdt0/6+JJeq6lOSPHj43QJgmiyvrCq6AMDY2y0cfWinB6rqwdbah5I8frhdAmCaLK+s5tylq1m/cTNJsrq2nnOXriaJgATAWNmtIMPbN29U1dvuemz50HsDwNhaXlnNqQuP5uGzj+TUhUezvLK6p+ddvHztVjDatH7jZi5evtZFNwFg33abOaott58+4DEApthBZn+ur63fUzsAjMpuM0dth9vb3QdgSh1k9ufYXO+e2gFgVHabOXpmVf3tbMwSbd5O//79nfYMgLFxkNmfpcWFO2adkqR39EiWFhcOrX8AcBh2C0c/lOSp29xOkh/upEcAjJ1jc72sbhOE9jL7s7nsTrU6AMbdwHDUWvt7w+oIAMOxn7LaB539OX1iXhgCYOwN3HNUVf9DVT23f7uq6tVV9TtV9d6qOjGcLgJwWDYLK6yurafldmGF3SrPnT4xn/Nnjmd+rpdKMj/Xy/kzxwUeAKbKbsvq/maS1/ZvvzTJFyR5TpITSb4/yZd21jMADt2gwgq7BR2zPwBMu92q1X2itXajf/srkry+tfZbrbWfSfJp3XYNgMOmrDYA7Gy3cPTJqnqgqj41yQuS/MyWx9RgBZgwymoDwM52C0ffnuRKkt9I8ubW2vuTpKr+XJJf77ZrABy2pcWF9I4euaNNWW0A2LBbtbqfqqrPTvLU1tpvb3nol5N8Tac9A+DQKasNADsbGI6q6k8m+c3W2n/s3//6JF+V5D8k+c7OewfAoVNYAQC2t9uyuv8jyX9Nkqr6s0kuJHl9kt9J8qpuuwYAADA8u5XyPtJa+3j/9tckeVVr7Y1J3lhV7+62awAAAMOz28zRkaraDFAvSPLolsd2C1YAAAATY7eA84YkP1tVH0uynuTnkqSqPicbS+sAAACmwm7V6r6rqt6W5IEk/7q11voP/YEkf6PrzgEAAAzLbtXqPr219q6721trv3rXNb/XRecAAACGZbc9R2+qqu+pqj9bVZ+22VhVz6mql1fV5SQv6raLAAAA3dttWd0LqurFSf7HJKeq6mlJPpHkWpJHkrxs8wwkAACASbZrxbnW2k8n+ekh9AUAAGBkdltWBwAAMBOcVQQwoZZXVnPx8rVcX1vPsblelhYXcvrE/Ki7BQATSzgCmEDLK6s5d+lq1m/cTJKsrq3n3KWrSSIgAcA+7WlZXVX94ap6Sv/286rqf66quW67BsBOLl6+disYbVq/cTMXL18bUY8AYPLtdc/RG5PcrKrPSfLPkzyc5F901isABrq+tn5P7QDA7vYajj7ZWvtEkr+Y5Ptaa38ryQPddQuAQY7N9e6pHQDY3V7D0Y2qemmSlyX5qX7b0W66BMBulhYX0jt65I623tEjWVpcGFGPAGDy7TUcfUOSP5Xku1prH6yqh5P8SHfdAmCQ0yfmc/7M8czP9VJJ5ud6OX/muGIMAHAA1Vq7tydUPS3JZ7XW3ttNl/bu5MmT7cqVK6PuBgAAMKaq6rHW2sm9XLvXanVvr6o/WFVPT/KeJK+pqn98kE4CAACMk72ec/QZrbXfrapvSvKa1tp3VNXIZ44A2J4DYgHg3u01HN1XVQ8k+ctJvq3D/gB0ahZCgwNiAWB/9lqQ4X9PcjnJr7XWfrmqnpPkA911C+DwbYaG1bX1tNwODcsrq6Pu2qFyQCwA7M+eZo5aa/8yyb/ccv/Xk3xVV50C6MKg0NDljMqwZ6scEAsA+7OncFRVz07yT5KcStKS/Nskf7O19kSHfQM4VKMIDQdZ4rbfUHVsrpfVbX4mB8QCwGB7XVb3miRvTnIsyXyS/6vfBjAxdgoHXYaG/S5xO8gSQAfEAsD+7DUc3d9ae01r7RP9r9cmub/DfgEculGEhv3OVh1k35ADYgFgf/Zare5jVfXfJ3lD//5Lk/xWN10C6MZmOBjm/p/9LnE76BLA0yfmhSEAuEd7DUffmOSVSb43G3uOfj7JN3TVKYCuDDs0LC0u3LHnKNnbbJV9QwAwfHtaVtda+1Br7Stba/e31p7ZWjud5EzHfQOYePtd4mbfEAAMX7XW9vfEqg+11h485P7ck5MnT7YrV66MsgsAnZmFA2sBoGtV9Vhr7eRert3rsrpt3+cAzwVgF/YNAcBw7bVa3Xb2N+UEAAAwhgbOHFXVf8r2IaiS2BUMAABMjYHhqLX21GF1BAAAYJQOsucIYKZ0USBB0QUAGB/CEcAeLK+s3nFe0eraes5duppko3DCfkLObq8JAAyXcASwBxcvX7vjINckWb9xMxcvX0uSfYWcQa8pHAHA8B2kWh3AzLi+tr5j+27BaT+vCQAMn3AEsAfH5rYv0HlsrrfvkDPoNQGA4ROOALZYXlnNqQuP5uGzj+TUhUezvLKaJFlaXEjv6JE7ru0dPZKlxYV9h5xBrwkADJ89RwB9eymQsFPRha3PS/YWcnZ7TQBguKq17c54PYQXrnp1kq9I8mRr7Y/2256e5MeTPJTkN5L85dbab295zp9M8q4kX9Na+8nd3uPkyZPtypUrh995YCaduvBoVrdZCjc/18s7zz5/4HOV5AaA8VRVj7XWTu7l2i5njl6b5JVJXr+l7WySt7XWLlTV2f79b02SqjqS5LuTXO6wTwA7OkiBhNMn5oUhAJhwne05aq29I8nH72p+SZLX9W+/LsnpLY/9jSRvTPJkV30CGESBBACYbcMuyPCs1tqHk6T//ZlJUlXzSf5ikh/c7QWq6hVVdaWqrnz0ox/ttLPAbFEgAQBm27hUq/u+JN/aWru524WttVe11k621k7ef//9Q+gaMCtOn5jP+TPHMz/XS2Vjr9H5M8ctlwOAGTHsanUfqaoHWmsfrqoHcnsJ3ckkP1ZVSfKMJC+uqk+01paH3D9gxtk7BACza9gzR29O8rL+7ZcleVOStNYebq091Fp7KMlPJvlmwQgAABimzsJRVb0hyS8kWaiqJ6rq5UkuJHlhVX0gyQv79wEAAEaus2V1rbWX7vDQC3Z53l89/N4AAAAMNi4FGQAAAEZKOAIAAIhwBAAAkEQ4AgAASCIcAQAAJBGOAAAAknRYyhtgVJZXVnPx8rVcX1vPsblelhYXcvrE/Ki7BQCMOeEI6Nwww8ryymrOXbqa9Rs3kySra+s5d+lqkghIAMBAwhHQqd3CymEHp4uXr916r03rN27m4uVrwhEAMJBwBHRqUFhJcuizPNfX1u+pHQBgk4IMQKcGhZXdgtN+HJvr3VM7AMAm4Qjo1KCw0sUsz9LiQnpHj9zR1jt6JEuLC/t+TQBgNghHQKcGhZUuZnlOn5jP+TPHMz/XSyWZn+vl/Jnj9hsBALuy5wjo1GYo2anowtY9R8nhzPKcPjEvDAEA90w4Ajq3U1jZLTgBAAyTcASMlFkeAGBc2HMEAAAQM0fAGDvsA2IBAAYRjmBCTXtwWF5ZPfQDYgEABrGsDibQZnBYXVtPy+3gsLyyOuquHZouDogFABhEOIIJNAvBoYsDYgEABhGOYALNQnDo4oBYAIBBhCOYQLMQHJYWF9I7euSOtsM4IBYAYCfCEUygWQgOp0/M5/yZ45mf66VD54S7AAAb+klEQVSSzM/1cv7MccUYAIDOqFYHE2gzIExztbrEAbEAwHAJRzChBAcAgMMlHAETadrPeQIAhk84ghkzDaHCAbEAQBcUZIAZMi2Hx87COU8AwPAJRzBDpiVUzMI5TwDA8AlHMEOmJVTMwjlPAMDwCUcwQw4SKpZXVnPqwqN5+OwjOXXh0ZEuxZuFc54AgOETjmCG7DdUjNteJQfEAgBdUK0OZsh+D48dtFfp9In5kVTAc84TAHDYhCOYMfsJFYP2KimrDQBMC8vqgF0N2qs0LRXwAACEI2BXg/YqTUsFPAAA4QjY1aACCMpqAwDTwp4jYE922qu0tLhwx56jRFltAGAyCUfAgey3Ah4AwLgRjoADU1YbAJgGwhFwyyjOKwIAGBfCEZAkzisCAGaeanVAkjivCACYecIRkGTnc4mcVwQAzArhCEiy87lEzisCAGaFcAQk2TivqHf0yB1tzisCAGaJggxAEucVAQAIR8AtzisCAGaZZXUAAAARjgAAAJIIRwAAAEmEIwAAgCTCEQAAQBLhCAAAIIlwBAAAkEQ4AgAASCIcAQAAJBGOAAAAkghHAAAASZL7Rt0BYHvLK6u5ePlarq+t59hcL0uLCzl9Yn7U3QIAmFrCEYyh5ZXVnLt0Nes3biZJVtfWc+7S1SQRkAAAOmJZHYyhi5ev3QpGm9Zv3MzFy9dG1CMAgOknHMEYur62fk/tAAAcnHAEY+jYXO+e2gEAODjhCMbQ0uJCekeP3NHWO3okS4sLI+oRAMD0U5ABxtBm0QXV6gAAhkc4gjF1+sS8MAQAMESW1QEAAEQ4AgAASGJZHUyl5ZVV+5UAAO6RcARTZnllNecuXb11iOzq2nrOXbqaJAISAMAAltXBlLl4+dqtYLRp/cbNXLx8bUQ9AgCYDMIRTJnra+v31A4AwAbhCKbMsbnePbUDALBBOIIps7S4kN7RI3e09Y4eydLiwoh6BAAwGRRkgCmzWXRBtToAgHsjHMEeTVJ57NMn5se2bwAA40o4gj3YrTz2JAUnAAC2JxzBHuxWHtu5QgAAk09BBtiDQeWxnSsEADAdhCPYg0HlsZ0rBAAwHYQj2INB5bGdKwQAMB3sOYI92K089tY9R4lzhQAAJpFwBHu0U3ns3YKTSnYAAJNBOIJDsFNw2q0EOAAA48OeI+iQSnYAAJPDzBF0aLdKdpbcAQCMDzNH0KFBlew2l9ytrq2n5faSu+WV1eF2EgCAJMIR3LK8sppTFx7Nw2cfyakLjx5KSBlUAtySOwCA8dJZOKqqV1fVk1X1vi1tT6+qt1bVB/rfn9Zv/7qqem//6+er6gu66hdsp6tZnNMn5nP+zPHMz/VSSebnejl/5nhOn5h3eCwAwJjpcs/Ra5O8Msnrt7SdTfK21tqFqjrbv/+tST6Y5M+11n67qr48yauSfHGHfYM7DJrFOegeoJ0q2R2b62V1myDk8FgAgNHobOaotfaOJB+/q/klSV7Xv/26JKf71/58a+23++3vSvLsrvoF2xnFLM6gJXcAAAzfsPccPau19uEk6X9/5jbXvDzJW4baK2beoMIJXRm05A4AgOEbq1LeVfXfZiMc/ZkB17wiySuS5MEHHxxSz5h2S4sLdxzWmgxnFmenJXcAAAzfsGeOPlJVDyRJ//uTmw9U1R9L8sNJXtJa+62dXqC19qrW2snW2sn777+/8w4zG8ziAAAw7JmjNyd5WZIL/e9vSpKqejDJpSR/pbX2q0PuEyQxiwMAMOs6C0dV9YYkz0vyjKp6Isl3ZCMU/URVvTzJh5J8df/yb0/ymUl+oKqS5BOttZNd9Q0AAOBunYWj1tpLd3joBdtc+01JvqmrvgAAAOxm2HuOAAAAxpJwBAAAEOEIAAAgiXAEAACQRDgCAABIIhwBAAAkEY4AAACSCEcAAABJhCMAAIAkwhEAAEAS4QgAACCJcAQAAJBEOAIAAEgiHAEAACQRjgAAAJIIRwAAAEmS+0bdARim5ZXVXLx8LdfX1nNsrpelxYWcPjE/6m4BADAGhCNmxvLKas5dupr1GzeTJKtr6zl36WqSCEgAAFhWx+y4ePnarWC0af3GzVy8fG1EPQIAYJwIR8yM62vr99QOAMBsEY6YGcfmevfUDgDAbBGOmDrLK6s5deHRPHz2kZy68GiWV1aTJEuLC+kdPXLHtb2jR7K0uDCKbgIAMGYUZGCq7KXogmp1AABsRzhiqgwqunD6xPytLwAAuJtwxEgd9rlDii4AALBf9hwxMptL4FbX1tNyewnc5h6h/VB0AQCA/RKOGJkuzh1SdAEAgP2yrI6R6WIJnKILAADsl3DEyByb62V1myB00CVwii4AALAfltUxMpbAAQAwTswcMTKWwAEAME6EI0bKEjgAAMaFZXUAAAARjgAAAJIIRwAAAEmEIwAAgCTCEQAAQBLhCAAAIIlwBAAAkEQ4AgAASCIcAQAAJEnuG3UHYD+WV1Zz8fK1XF9bz7G5XpYWF3L6xPyouwUAwAQTjpg4yyurOXfpatZv3EySrK6t59ylq0kiIAEAsG/C0RSa9lmVi5ev3QpGm9Zv3MzFy9em6ucEAGC4hKMpMwuzKtfX1u+pHQAA9kJBhikzaFZlWhyb691TOwAA7IVwNGUOMquyvLKaUxcezcNnH8mpC49meWX1sLt3KJYWF9I7euSOtt7RI1laXBhRjwAAmAaW1U2ZY3O9rG4ThHabVZmk5Xib/ZnmfVUAAAyfcDRllhYX7gg5yd5mVSatyMHpE/Nj2S8AACaXcDRl9jurosgBAACzTjiaQvuZVdnvcrwuTXtJcgAAxouCDCQZvyIHm3ugVtfW03J7D9S4FokAAGDyCUck2ZhtOn/meObneqkk83O9nD9zfGQzNbNQkhwAgPFiWd0E6mq52TgVObAHCgCAYTNzNGFmZbmZg14BABg24WjCzMpys3HbAwUAwPSzrG7CzMpyMwe9AgAwbMLRhBnHkttdGac9UAAATD/L6iaM5WYAANANM0cTxnIzAADohnA0gSZpuVlXZccBAOCwCUd0ZrPs+GZ1vc2y40kEJAAAxo49R3RmVsqOAwAwHcwczZhhLnOblbLjAABMBzNHM2Rzmdvq2npabi9zW15Z7eT9diovPo1lxwEAmHzC0Qw5yDK35ZXVnLrwaB4++0hOXXh0T4FK2XEAACaJZXUzZL/L3HYrrLDTUj1lxwEAmCTC0Qw5NtfL6jZBaLdlbrvNOA0KTpNUdhwAgNlmWd0M2e8yt0EzTirSAQAwLYSjGXL6xHzOnzme+bleKsn8XC/nzxzfdWZnUGEFFekAAJgWltXNmP0sc1taXLhj6Vxye8bp4uVr+1qqBwAA40Y4Yle7FVbYKTgBAMAkEY7Yk51mnFSkAwBgWghHHJiKdAAATAMFGQAAACIcAQAAJBGOAAAAkghHAAAASYQjAACAJMIRAABAEuEIAAAgiXAEAACQRDgCAABIIhwBAAAkEY4AAACSCEcAAABJkvtG3QG2t7yymouXr+X62nqOzfWytLiQ0yfmR90tAACYWsLRGFpeWc25S1ezfuNmkmR1bT3nLl1NEgEJAAA6IhyN0E6zQxcvX7sVjDat37iZi5evCUcAANAR4WhEBs0OXV9b3/Y5O7UDAAAHpyDDiAyaHTo219v2OTu1AwAAB9dZOKqqV1fVk1X1vi1tT6+qt1bVB/rfn9Zvr6r6/qp6vKreW1V/vKt+jYtBs0NLiwvpHT1yR3vv6JEsLS4Mo2sAADCTupw5em2SF93VdjbJ21prz03ytv79JPnyJM/tf70iyT/rsF9jYdDs0OkT8zl/5njm53qpJPNzvZw/c9x+IwAA6FBne45aa++oqofuan5Jkuf1b78uyduTfGu//fWttZbkXVU1V1UPtNY+3FX/Rm1pceGOPUfJnbNDp0/MC0MAADBEwy7I8KzNwNNa+3BVPbPfPp/kN7dc90S/beLD0U4V6TaDj7OMAABgPIxLtbrapq1te2HVK7Kx9C4PPvhgl306sN3OKzI7BAAA42PY1eo+UlUPJEn/+5P99ieSfNaW656d5Pp2L9Bae1Vr7WRr7eT999/faWcPalBFOgAAYLwMOxy9OcnL+rdfluRNW9q/vl+17kuS/M407DdyXhEAAEyOLkt5vyHJLyRZqKonqurlSS4keWFVfSDJC/v3k+Snk/x6kseT/FCSb+6qX8PkvCIAAJgcXVare+kOD71gm2tbkm/pqi+jsltFOgAAYHyMS0GGqaQiHQAATA7hqGMq0gEAwGQQjg7BTmcZAQAAk0M4OqDdzjICAAAmw7BLeU8dZxkBAMB0EI4OyFlGAAAwHYSjA3KWEQAATAfh6ICWFhfSO3rkjjZnGQEAwORRkOGAnGUEAADTQTg6BM4yAgCAyWdZHQAAQIQjAACAJMIRAABAEuEIAAAgiXAEAACQRDgCAABIIhwBAAAkEY4AAACSCEcAAABJhCMAAIAkwhEAAEAS4QgAACCJcAQAAJBEOAIAAEgiHAEAACQRjgAAAJIIRwAAAEmEIwAAgCTCEQAAQBLhCAAAIElSrbVR92HfquqjSf7DqPsxRp6R5GOj7sQMMM7dM8bDYZyHwzh3zxgPh3EeDuN8+D67tXb/Xi6c6HDEnarqSmvt5Kj7Me2Mc/eM8XAY5+Ewzt0zxsNhnIfDOI+WZXUAAAARjgAAAJIIR9PmVaPuwIwwzt0zxsNhnIfDOHfPGA+HcR4O4zxC9hwBAADEzBEAAEAS4WjsVNWnVtUvVdV7qur9VfX3+u0/V1Xv7n9dr6rlfvvzqup3tjz27Vte60VVda2qHq+qs1vaH66qX6yqD1TVj1fVpwz/Jx29qjpSVStV9VP9+9uOS1U9pX//8f7jD215jXP99mtVtbilfduxn0XbjPOP9sfmfVX16qo62m/3WT6Abcb5tVX1wS3j+YX99qqq7++P5Xur6o9veY2X9cfyA1X1si3tf6Kqrvaf8/1VVcP/CUdvmzH2e/mQVdVv9D9r766qK/22p1fVW/tj89aqelq/3Wd5n3YY54tV9e/7Y/mvqmqu3/5QVa1v+Tz/4JbX2XY8d/ozmzU7jPN3VtXqlvF88Zbr7+nvFH5vdKS15muMvpJUkk/v3z6a5BeTfMld17wxydf3bz8vyU9t8zpHkvxakuck+ZQk70nyef3HfiLJ1/Zv/2CS/2nUP/eIxvpvJ/kXm+O307gk+eYkP9i//bVJfrx/+/P64/qUJA/3x/vIoLGfxa9txvnF/c95JXnDlnH2WT7ccX5tkr+0zXUvTvKW/vh/SZJf7Lc/Pcmv978/rX/7af3HfinJn+o/5y1JvnzUP+84jPFdj/m9fDhj/BtJnnFX2z9McrZ/+2yS7+7f9lk+3HH+siT39W9/95ZxfijJ+3Z4nW3Hc6c/s1n72mGcvzPJ/7rNtff8dwq/N7r5MnM0ZtqG3+vfPdr/urUxrKqemuT5SZZ3eakvSvJ4a+3XW2v/NcmPJXlJ/191np/kJ/vXvS7J6UP8ESZCVT07yZ9P8sP9+4PG5SX9++k//oL+9S9J8mOttf/SWvtgksezMe7bjn33P9X4uXuck6S19tP9z3nLxv9Yn73Ly/gs72K7cR7gJUle3/8jeFeSuap6IMlikre21j7eWvvtJG9N8qL+Y3+wtfYL/T+z12cGx3nQGPu93Lmtv4Pv/t3ss3xIWmv/urX2if7dd2WX3827jOdOf2bs7J7+TuH3RneEozHUX7rx7iRPZuMX/C9uefgvJnlba+13t7T9qdpYhveWqvr8ftt8kt/ccs0T/bbPTLK25RfgZvus+b4kfyfJJ/v3B43LrbHsP/47/et3GuOd2mfR3eN8S20sp/srSf7vLc0+y/uz0zh/V3+JzPdW1VP6bff6uZ3v3767fdbs+FmO38uHqSX511X1WFW9ot/2rNbah5Ok//2Z/Xaf5f3bbpy3+sZszARterg2lpT+bFV9ab9t0Hju9Gc2a3Ya57/e/9386i1LDu/18+z3RkeEozHUWrvZWvvCbPyrzRdV1R/d8vBLs7EUadO/S/LZrbUvSPJPcvtfLrdbR90GtM+MqvqKJE+21h7b2rzNpW2Xx4zxADuM81Y/kOQdrbWf69/3Wd6HAeN8LskfSfIns7G86Fs3n7LNy/g8D7CHz7Lfy4fnVGvtjyf58iTfUlV/dsC1Psv7t+M4V9W3JflEkh/tN304yYOttRPpLy2tqj8Y47kX243zP0vyh5N8YTbG9nv61/o8jwnhaIy11taSvD3Ji5Kkqj4zG9Orj2y55nc3l+G11n46ydGqekY2/gXhs7a83LOTXE/ysWwsPbjvrvZZcirJV1bVb2Rjevr52fhX4Z3G5dZY9h//jCQfz85jvFP7rPl941xVP5IkVfUdSe7Pxv9ok/gsH8C249xa+3B/udF/SfKabPzuSO79c/tE7lxeM4vjPOiz7PfyIWqtXe9/fzLJv8rG2H6kv4RrcynXk/3LfZb3aYdxTm0Ur/iKJF/XXyqX/jKv3+rffiwb+18+N4PHc6c/s5my3Ti31j7S/0fwTyb5oez/d7PfG105jI1Lvg7vKxt/YZzr3+4l+bkkX9G//9eSvO6u6/9Qbp9X9UVJPpSNf024LxubUB/O7Q18n9+/7l/mzg183zzqn3uE4/283N7Avu24JPmW3FmQ4Sf6tz8/d26e/PVsbJzccexn9euucf6mJD+fpHfXNT7LhzvOD/S/VzbC/4X+/T+fOzex/1K//elJPpiNDexP699+ev+xX+5fu7np+sWj/lnHYYz79/1ePryx/bQkT91y++ez8Y+DF3Pn5v5/2L/ts3y44/yiJL+S5P67rr8/yZH+7eckWd1tPHf6M5ulrwHj/MCWa/5WNvYZJfv4O4XfGx392Y26A77u+gNJ/liSlSTvTfK+JN++5bG3J3nRXdf/9STv7//H8q4kf3rLYy9O8qvZ+Feeb9vS/pxsbIR/vP8f1lNG/XOPcLyfl9t/mdx2XJJ8av/+4/3Hn7Pl+d/WH99r2VL1aKexn9Wvu8b5E/1xeXf/69v77T7LhzvOjya52v898iO5XQWzkvzT/lheTXJyy/O/sT+Wjyf5hi3tJ/uv82tJXpn+X/xn8Su/Pxz5vXx4Y/uc/pi9pz9+39Zv/8wkb0vygf73zb+Y+ywf7jg/no29LZu/mzf/UfCrtnye/12Sv7DbeO70ZzZLXwPG+f/sf17fm+TNuTMs3dPfKfze6OZr80MMAAAw0+w5AgAAiHAEAACQRDgCAABIIhwBAAAkEY4AAACSCEcAQ1dVn1lV7+5//ceqWt1y/1P2+BqvqaqFXa75lqr6usPp9bavf6aq/khXr99/jz9QVf+mqj69qu6rqptbxurdVbV0iO/1OVX17sN6vX28f+fj2X+fv1tVX3NX29dX1dWqem9VvbOqjvfbP7WqfraqjnTdL4BxcN/ulwBwmNrGafNfmCRV9Z1Jfq+19o+2XlNVlY0zQz65w2t8wx7e558evLcDnUnyyST/vsP3+AtJrrTWfq9/Evx/aq19YYfvN0rDGM8keWE2zgfa6teSfGlrba2q/kI2DpQ81Vr7/6rqHUn+UpIf77hfACNn5ghgTPRnLt5XVT+YjcMWH6iqV1XVlap6f1V9+5Zr/21VfWF/NmWtqi5U1Xuq6heq6pn9a/5BVf0vW66/UFW/VFXXqupP99s/rare2H/uG/rv9fvCR1VdrKpf6c8sfHdVfWk2Dib83v4MzkNV9dyqulxVj1XVO6rqc/vP/ZGq+mdV9XNV9atV9eX99uNV9cv957+3qp6zzbB8XZI37WHsntjy8/3i5mtV1cP9maf3VtVbq+rZ/fY/VFVv6re/p6q+uP9S91XVP++P91uq6lP71/+t/s//nqr6kT39gQ7u70jGs6rmkqS19vGt7a21d7bW1vp335Xk2VseXs7GnwPA1DNzBDBePi/JN7TW/lqSVNXZ1trH+7Mm/6aqfrK19it3Peczkvxsa+1sVf3jJN+Y5MI2r12ttS+qqq9M8u1JXpTkbyT5j621r6qqL8hGKLvzSVXPysZf3D+/tdaqaq4/w/DTSX6ytbbcv+7fJPmm1tqvVdWpJK9M8mX9l/msJH8uyXOT/ExVfU6Sb07yj1prP15VT0lS2/T5VJK/uuX+U+vOpW//oLX2k/3bv93/+b4xyT9OcjrJDyT54dbaj1bVK5J8XzZmQf5pkre21l7ZH9v/JskzkywkeWlr7WpVXeq/xo8l+TtJPru19l83A8ZdY/R5Sf7FNv1PNmZk/tOYjOeXJfmZHfq56eVJ3rLl/nuSfMkuzwGYCsIRwHj5tdbaL2+5/9Kqenk2fl8fy0Z4ujscrbfWNv8y+1iSL93htS9tueah/u0/k+S7k6S19p6qev82z/t4NpZ7/VBVPZLkp+6+oB8YviTJG6tu/Z186/9jfqK/RPBaVf1mNv5S//NJ/req+uwkl1prj2/z3k9trf2/W+4PWlb3hv73H83tcPjF+f/buYMQq6owgOP/z0QsnBECF5KLEKZFiBhGSzEJNAhCojSxrSYEYeLGVRAtWkoLkTZRohs3UgOROEEUlIiUk22apdJCIWqK0JTPxTk373veV/Nq4I3P/w8uHN495957DrO435zvfvBCbX8EvFPbW4HdAJl5C/it7rjNZeZs7dNep8vAiYg4Q9lJ6VED1oWm+41yPXcAxwY9WEQ8B7xG+bto5nYrIjIiHs7MPxc4R0m6L5lWJ0lLyx9NIyKmgDeBbZm5EfgMWNkx5marfZvB//i60dGna3ehR2b+BTxNCQpeAqY7ugVwPTM3tY4N7cvce9n8GNhZn+tsRGzpuG7nN1eDHnWIvoP632i12+u0nfIdzjPAhegrUBART0ZvoYj2MdFz09Gu52ZK0HfvDUs65XHgxcz8pe/0CnrXRpLGksGRJC1dk8A8ZVdjLeUFfbF9BbwC5ZsVys5Uj/pyP5mZnwIHgafqqXlgAqC+TP8cETvrmGU1Ta/xchRPUFLCfoqI9Zk5l5lHKQHCxo7nm4uIxxc4l6YC26vA17X9TTM/YC/wZW1/ATSpiw9FxOSgi9ZAaF1mzgCHgTWUNLy/ZeaPfYFM+5jvu95I1rOOn+0q8lHX+DSwp3/HqaYBXh1UHESSxonBkSQtXRcpKXQ/AB9w94V/Mb0PPBYRl4BD9V6/9vVZDUxHxPfADPBW/f0UcKQpIEBJU3u99rvM3XQ2gDlKYPIJsC8zbwJ7auGD74D1QFehg2lKClxjom9X5t3WuUci4jxwoM4F4A1gX53fLkow0vy+PSJmgQvAP5XQXg6crNe4CLzXH/AMaVTr+Txl97HL28CjwPF6/29b556le3dLksZOZA6bhSBJGhe1GMHyWrJ5CvgcmKrf4SzWPU7QKjQw5Nh1lIIKO/6l3xVgQ6vi2tj6r+sZETPArsy8NuS4M8ChAd8wSdJYsSCDJD3YVgHnapAUwP7FDIz+r8y8EhEfRsSqzPx91M9zP8vMbcOOqVXvThsYSXpQuHMkSZIkSfjNkSRJkiQBBkeSJEmSBBgcSZIkSRJgcCRJkiRJgMGRJEmSJAEGR5IkSZIEwB3H6s9ySnVq4gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1008x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.rcParams['figure.figsize'] = [14, 10]\n",
    "\n",
    "loss_values = [ev['loss'] for ev in evaluations]\n",
    "training_steps = [ev['global_step'] for ev in evaluations]\n",
    "\n",
    "plt.scatter(x=training_steps, y=loss_values)\n",
    "plt.xlabel('Training steps (Epochs = steps / 2)')\n",
    "plt.ylabel('Loss (SSE)')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "WARNING:tensorflow:Layer dnn is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because it's dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from tf_wx_model\\model.ckpt-56400\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n"
     ]
    }
   ],
   "source": [
    "pred = regressor.predict(input_fn=wx_input_fn(X_test,\n",
    "                                              num_epochs=1,\n",
    "                                              shuffle=False))\n",
    "predictions = np.array([p['predictions'][0] for p in pred])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The Explained Variance: 0.26\n",
      "The Mean Absolute Error: 7.76 degrees Celcius\n",
      "The Median Absolute Error: 8.39 degrees Celcius\n"
     ]
    }
   ],
   "source": [
    "print(\"The Explained Variance: %.2f\" % explained_variance_score(\n",
    "                                            y_test, predictions))\n",
    "print(\"The Mean Absolute Error: %.2f degrees Celcius\" % mean_absolute_error(\n",
    "                                            y_test, predictions))\n",
    "print(\"The Median Absolute Error: %.2f degrees Celcius\" % median_absolute_error(\n",
    "                                            y_test, predictions))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}